Help talk:IPA/English/Archive 10

This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Archive 5

←

Archive 8

→

Audio

Latest comment: 13 years ago6 comments3 people in discussion

This article really really really needs an audio file to illustrate each of the sounds. 76.85.196.138 (talk) 04:40, 5 April 2010 (UTC)

On second thought, the entire use of an IPA on wikipedia is a tedious obfuscation compared to just inserting a widget that plays a sound file. At best, the IPA gives you a crude approximation of the pronunciation with some considerable effort at decoding it.76.85.196.138 (talk) 04:59, 5 April 2010 (UTC)

One can actually do both. If we were to have a policy of using sound files to indicate pronunciation, we would have to create quite a few thousand sound files. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 06:12, 5 April 2010 (UTC)

There's also the problem of whose dialect we use. kwami (talk) 19:57, 20 April 2010 (UTC)

Where the pronunciation differs by dialect, the IPA representation will also differ. So that problem is not applicable to comparing the merits of sound files and IPA. Sound files are more precise and accessible. IPA is too reductionist. Making a few thousand of these sound files would be a good job for a new WikiProject. They could be added to articles in addition to or instead of IPA. 76.85.196.138 (talk) 19:46, 1 June 2010 (UTC)

No, you evidently haven't read any of this. The IPA will not differ, but the sound files will. So sound files cannot explain the IPA, unless we can somehow link to sound files in each reader's dialect. — kwami (talk) 20:57, 1 June 2010 (UTC)

More discussion needed

Latest comment: 13 years ago163 comments11 people in discussion

Having read through this page and some of the archives it seems to me that a very far reaching decision has been taken by a small number of editors.

We have a very difficult problem how to give pronunciations for a language where there are many different accents spread across many countries. At one extreme there is much commonality in English. Speakers using most English accents can usually understand one another. On the other hand no two people's pronunciation is exactly the same. Martin Hogbin (talk)

I have not seen much discussion of alternatives (although I admit I have not read all the archives). Martin Hogbin (talk) 22:43, 15 May 2010 (UTC)

This has been the consensus for several years. It's not just the editors who came up with the compromise in the first place (and I don't think any of us got exactly what we wanted. I know I didn't: I wanted something more phonemic), but also the many others who have used it. Since it's now found on 10,000 articles, and there have been few complaints (and those mostly from newbies who haven't read the key), I don't think anyone can say that there isn't general consensus on this convention. There is far more complaint about using the IPA in the first place than about this key.

What else would you suggest? If we choose a particular national or dictionary standard instead, then there will be accusations of cultural imperialism. (And should we use RP, because English started in England, or GA, because the US is the largest anglophone country?) If we counter this by adding the pronunciation in the national standard of every English-speaking country which has enough editors to demand that they too be represented, then the entries will quickly become unwieldy, and we'll get complaints that WP is not a dictionary. — kwami (talk) 23:19, 15 May 2010 (UTC)

The wrong vehicle

I support the use of IPA pronunciation in WP but, if we are going to go down the route of inventing a new general English pronunciation, I suggest IPA is the wrong vehicle. The IPA was created to represent different sounds accurately, specifically not to require interpretation. To use the IPA in a situation where readers are expected to interpret it according to their own accent is a significant diversion from it original purpose. In other words, if someone was going to try to create a system for accent-free representation of English, the IPA would not be the way to do it. Martin Hogbin (talk) 23:16, 15 May 2010 (UTC)

This is the difference between broad and narrow transcriptions, between phonetic and phonemic (or in this case, diaphonemic). The IPA has long been used for both. /r/ is not the same thing as [r]. We also have very broad consensus that the IPA, as the international standard, is the way to go. The only exceptions are Americans whose provincial education did not prepare them for the outside world, rather like demanding that we jettison the metric system as being unworkable because they aren't familiar with it. — kwami (talk) 23:21, 15 May 2010 (UTC)

For a concept that our transcription hinges on, the article diaphoneme is atrociously shallow. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 03:44, 16 May 2010 (UTC)

That's a convenient label, but I'm not sure it's the best. "Diaphoneme" is sometimes (and I think originally) presented as something cognitively real, part of the speaker's lexicon, which is of course nonsense apart from those of us who are bi-lectal. Someone who speaks both Cockney and RP may have them cognitively unified diaphonically in their mental lexicon (or maybe not), but these RP vowel distinctions I'm transcribing here on WP have nothing to do with my own speech, and are no more real in my brain than the /r/ in "York" is for locals. So I don't think "diaphonemics" offers any kind of theoretical justification. The word is simply a label for an orthographic convenience, because the WP system isn't exactly phonemic, and calling it that as we used to could cause some confusion. The AHD system is also like this, but I've never seen a name for it. It's like the asterisked forms you see in language reconstructions: are those phones "phonemes"? "diaphonemes"? "eophonemes"? — kwami (talk) 06:37, 16 May 2010 (UTC)

That's something that can be researched. It may not be cognitively real, but beefing up that article and representing the work of scholars who have done in the past what we do with {{IPA-en}} are the best ways to avoid accusations of unwarranted original research. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 08:26, 16 May 2010 (UTC)

I agree that the IPA is the way to go for precise transcription of sounds (as you see, I am not a linguist but I hope to learn the correct technical terms as I go) but it is ill suited to the purpose that editors here seem to have in mind. If we are going to use the IPA then why not stick to giving RP, GA, and any other accents required. Why has this practice been abandoned? What editors here are attempting to do is more that a broad transcription of IPA. I understand the objective here to be to give a transcription that represents in some way the pronunciation in more that one accent. What it actually does is give a transcription in an accent that nobody uses.Martin Hogbin (talk) 08:47, 16 May 2010 (UTC)

The IPA is a code. Substituting some other code letter by letter makes no difference in content whatsoever. WP has decided to primarily use the IPA, and jettisoning that for some provincial system like AHD for zero gain in content makes no sense to me. As for why we abandoned multiple transcriptions, would you really want us to list the Australian, New Zealand, South African, Jamaican, Irish, Scottish, RP, GA, Canadian, and Southern US pronunciations of "chamois"? What's the point? Yet if you were to decide that any one of those nations was not important enough to warrant their own transcription, you'd be "oppressing" them. — kwami (talk) 10:42, 16 May 2010 (UTC)

The IPA is indeed a code but it is a code with a purpose, which is to accurately represent speech. The IPA lets us represent all the multiple pronunciations of 'chamois' accurately and unambiguously. Up until this policy decision, that is exactly what it was used for on WP.

One option is to allow multiple pronunciations covering different accents, this would not be easy but that does not mean that we should not do it. WP as a medium allows solutions that are not applicable to a written encyclopedia, user settings, for example. My point concerning multiple transcriptions is not that we must do this, just that we have not considered other options seriously enough.

In the case of your example, how does the use of a special version of the IPA solve the problem? Martin Hogbin (talk) 11:26, 16 May 2010 (UTC)

The IPA is designed to give accurate phonetic transcription. It is also used to represent the underlying phonemic ("broad") representation of sounds, which is slightly more abstract but is supposed to be cognitively real. The word cat, for example, ends in a glottal stop [kʰæʔ] for many speakers. However, because of the way the glottal stop patterns and because the word cats is [kʰæts] with a [t] in that same place, it is understood that the glottal stop is a contextual realization of an underlying /t/ phoneme. Our system goes one step further to represent underlying representations of multiple dialects, which isn't cognitively real but I believe has been done elsewhere. An outline of English structure by Trager and Smith (1951) might be a good start. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 17:33, 16 May 2010 (UTC)

Thanks for the book recommendation. Perhaps you could explain how the use of your scheme helps with the problem that kwami has identified above. Martin Hogbin (talk) 18:38, 16 May 2010 (UTC)

Certainly. The scheme attempts to encode the phonemic contrasts of all the relevent dialects, including those that aren't common to all. This way, one can take the transcription given and apply phonemic mergers of a particular dialect to represent said dialect. For example. The transcription /njuː ˈjɔrk/ for New York has a palatal approximant after the n. If you, like me, speak a dialect that features yod-dropping, then you know to ignore the /j/ after alveolar sounds (/s l z θ t d n/; speakers of dialects that don't feature yod-dropping make meaningful contrasts between words like loot and lute, and it's easier to ask readers to apply a merger than to undo one. At the same time, there are dialects that don't pronounce an r the syllable coda so that speakers of those dialects can ignore the /r/ in that transcription. By doing it this way, we can provide one transcription that covers all dialects.

By the way, just to clarify, I haven't read that book so I don't know how good it is. I'm just starting to do research on information related to diaphoneme and that name came up. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:56, 16 May 2010 (UTC)

I was actually asking how the scheme would help with the example kwami gave above, but in any case it seems crazy to me. How is the reader to know to ignore the /j/ after alveolar sounds when it is specifically shown in the IPA transcription? This requires even more knowledge from the reader than simply knowing the IPA. On the other hand, with a suitable respelling scheme, which in this case could just have 'new' for the first word (with an appropriate key of course) the variation in pronunciation between yod-droppers and non-yod-droppers would be completely intuitive and no alphabets get hurt in the process. Martin Hogbin (talk) 20:08, 16 May 2010 (UTC)

Oh, my bad. I missed the example in Kwami's poist. Chamois is not a good example because the differences in pronunciation between UK and US go beyond phonemic differences. There are a handful of such words and, in those cases, our diaphonemic transcription doesn't work. However, let's take another example. Imagine if our article reindeer began like this:

The reindeer, (Received Pronunciation [ˈreɪndɪə]; General American [ˈreɪndiːr]; Canadian English [ˈreːndir]; Australian English [ˈræɪndɪə]; New Zealand English [ˈræendɪə]), Rangifer tarandus, also known as the caribou (General American [ˈkɛrəˌbu]; Canadian English [ˈkɛrəˌbʉ]) in North America, is a deer from the Arctic and Subarctic, including both resident and migratory populations...

Under our diaphonemic system, it would look like this.

The reindeer (/ˈreɪndɪər/ Rangifer tarandus), also known as the caribou /ˈkær[invalid input: 'ɨ']ˌbuː/ in North America, is a deer from the Arctic and Subarctic, including both resident and migratory populations...

It's simpler, easier to read, consistent, and easier to manage. This very page describes how to read the transcription. We can't expect everyone to read the key, but if anyone is confused by the transcription system, then they'll be prompted to visit here and read "understanding the key." As far as knowledge of the IPA goes, using multiple systems actually requires a more thorough knowledge of the IPA (see Kwami's post above dated February 28). — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 00:14, 17 May 2010 (UTC

Has it still never occurred to the IPA experts here that most people looking at an IP transcription are not obsessed with linguistics? Has it never occured to the IPA specialists here that the IPA uses a character set that more or less looks like the Roman alphabet? Has is never occured to the IPA 'specialists' that non linguists will probably tend to voice an IPA character in the way way they pronounce a letter from the Roman alphabet, and that they most likely will not bother to look up something in a key that is going to baffle them even more with science? How many of the IPA specialists here have actually tried teaching the IPA or better still, a foreign language? In the Wikipedia, the IPA and all its keys and explanations should try to appeal to a wider, non specialised audience, rather than the IP editors writing for themselves or for each other in order to show off their knowledge of the subject.--Kudpung (talk) 00:28, 17 May 2010 (UTC)

Are you saying that a reader who sees /ˈreɪndɪər/ might pronounce it as if it's spelled reindier? Do you see this as more or less likely than a nonspecialist reader seeing /ˈreɪndɪər/ and actually pronouncing the /r/ in the syllable coda despite speaking a non-rhotic dialect? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 04:54, 17 May 2010 (UTC)

That is exactly what the reader should do. That is the whole point of the IPA. As I have already suggested, a respelling scheme based on the pronunciation of well-known English words solves this problem much better. I see on the project that you are linking IPA symbols to standard lexical sets. This is crazy and completely against the spirit of the IPA. If you really want to do this kind of thing it would be better to devise a completely new set of symbols rather than abuse the IPA in this way. Actually, the answer is much simpler. Just use the well though out lexical sets that you already have directly as the basis of a respelling scheme. You are halfway there. Martin Hogbin (talk) 09:58, 17 May 2010 (UTC)

My question was directed at Kudpung, who has now said that readers of non-rhotic dialects would be prompted to pronounce a word-final r and that they would also be prompted to read IPA transcription as if it were not written in IPA at all. These seem contradictory issues to me and require different solutions. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:11, 17 May 2010 (UTC)

If I were going to Dublin I would not start from here

The first thing we need to do is to consider whether we should be devising some diaphonemic (is that a real word?) transcription scheme for English. The alternative is to give multiple IPA pronunciations covering the maim English accents. There is a good case for doing this, which is that it gives the most and the best quality information to our readers, and I would suggest that there should be very wide consensus before changing this practice.

If it is considered that a diaphonemic transcription scheme would be beneficial for WP then I would suggest two things, firstly that it is used, at least initially, along side accurate regional transcriptions using the IPA.

Secondly, the natural medium for transcribing English in such a way that readers will hear the words in their own accent would be a respelling scheme based only on the Latin alphabet. All English words can be spelled using the Latin alphabet and readers already find it natural to interpret the pronunciation differently according to their individual accent. To devise such a scheme would be a significant undertaking and it would first be worth checking whether this idea has been proposed before, but if we are going to try to do this job I believe we should do it properly. The current scheme is a compromise at best and will annoy readers at both ends of the spectrum. Those not familiar with the IPA will still dislike it and many regular users of the IPA may see it as an abuse of a system that was designed to avoid ambiguity.

Sorry to pour cold water on your hard work folks but I think you need to stop and think about what exactly you are trying to achieve here and what is the best way to do that. Martin Hogbin (talk) 08:47, 16 May 2010 (UTC)

You do realize that we already allow a pronunciation respelling to go alongside the IPA, right? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 08:58, 16 May 2010 (UTC

Yes, and I support that. My point is that if we are intending to develop a diaphonemic transcription scheme this should be done by using a WP diaphonemic respelling scheme rather than by mangling the IPA. The IPA should be used to give accurate and specific pronunciations only. Martin Hogbin (talk) 11:12, 16 May 2010 (UTC)

Can I just repeat what you said (in support of it)?: The IPA should be used to give accurate and specific pronunciations only. --Kudpung (talk) 15:06, 16 May 2010 (UTC)

What do we use for all the phonology articles, then, if we shouldn't use the IPA for phonemes? A phoneme, after all, is not specific, and generally does not correspond to actual pronunciation. — kwami (talk) 02:21, 17 May 2010 (UTC)

Not exactly but very closely. Here we are talking about significant and easily noticed differences. Most people can easily hear the difference between GA and RP even if they cannot explain exactly what the difference is. In your scheme you are asking readers to intentionally misread the IPA.

It is not even clear how the readers are expected to get to the correct pronunciation. Are they expected to look at the IPA representation and change it according to a set of rules and then read the new IPA spelling, or are they supposed to listen in their heads to the rather odd mid-Atlantic accent represented by the actual IPA transcription given and imagine how this would sound in their own accent? The IPA for English gives little clue on how to do this at all. Indeed, some readers may not even know how their own accent is described.

I fully understand and support the intention to give multiple pronunciations easily and conveniently but you simply have not thought this one through properly. There are several other options, including technical solutions, that you have not even considered. If you want to use a written scheme to solve the multiple accent problem then a respelling scheme based on English spelling of common and well-know words is clearly the way to do this, as people automatically render such spellings into their own accent. I am not in any way against using the IPA, but it should only be used for its proper purpose of precisely representing speech. Indeed, I still support its use for giving accurate pronunciations in specific accents, for example local pronunciation of geographic names. Using repelling may seem a backward and unscholarly step to some but it is the way that best suits the purpose we have of giving the reader a simple indication of how a given word would sound in their own accent. We can never hope to do this precisely because we can never know the exact accent of an individual reader. On the other hand each reader does know how they pronounce common English words.

My suggestion is also far less contentious in that, because WP policy already allows respelling to be used in addition to IPA we can simply devise and recommend a WP multi-accent respelling scheme that could be optionally added to articles. I can assure you that this will go down much better than (apparently) telling people from Worcester how to pronounce the name of their home town.

I may not have the perfect solution but I would ask you all to stop and think before demanding that everyone in WP follows this new policy. Martin Hogbin (talk) 09:45, 17 May 2010 (UTC)

This isn't a new policy. It isn't a policy at all, it's a guideline, but it isn't even a new guideline. It's been in place for years, and all of your arguments are the same ones that were brought up and taken into consideration when the guide was being worked out. And if you think that the only "proper purpose" of the IPA is "precisely representing speech" and "giving accurate pronunciations in specific accents" then you severely misunderstand the IPA. The IPA can be used to represent speech phonetically, but it can – and always has – also been used to represent speech broadly and phonemically and, yes, diaphonemically. You say "If you want to use a written scheme to solve the multiple accent problem then a respelling scheme based on English spelling of common and well-know[n] words is clearly the way to do this, as people automatically render such spellings into their own accent", but that simply isn't the case. A respelling scheme based on English orthography will be no easier to learn or to recognize or to convert to one's own pronunciation than the IPA is. We've tried it; we even have a pronunciation respelling key which we use alongside IPA, and it's no simpler and no more intuitive than IPA is. Both systems require learning how to match up a set of symbols to a smaller set of phonemes. And using the respelling key will certainly not solve the problem we're currently having at Talk:Worcester: non-rhotic speakers will reject WOO-stər just as much as they reject /ˈwʊstər/ and for the same reasons; and yod-dropping speakers will reject nyoo-MEK-sih-koh just as much as they reject /njuː ˈmɛksɨkoʊ/ and for the same reasons. +An gr 16:10, 17 May 2010 (UTC)

If this is not a policy it certainly is being used like one. You yourself said on the Worcester page, 'Responding to the RFC: WP:IPA for English is quite clear; the transcription should be /ˈwʊstər/'. I have had a quick look through some of the talk page archives and have not yet found a proper discussion of all the possible alternatives to the current scheme. Perhaps you could direct me to this discussion.

I admit that I am not an expert in the IPA. Can you give me examples of other places where a single IPA symbol is used to represent such a wide range of sounds as we are doing here here?

The current pronunciation respelling key might be a good place to start but it is not exactly what I meant. Firstly, it does not have the stated aim of giving a pronunciation using a single transcription to represent many different accents. The key does not do this either since it has IPA values for the respelling symbols. Well-known ordinary English words are automatically transcribed into the accent of the reader, this is how reading works. In the example you give, use of the English word 'new' solves the problem completely.

I do not want to deprecate the hard work that you and others have put into this page but I strongly suggest that some kind of review is called for before you force the results (as you have been doing) onto other WP articles Martin Hogbin (talk) 17:24, 17 May 2010 (UTC)

What I'm confused by, Martin, is that this page-both in the "understanding the key" section and at the top of the talk page here-address some of your concerns. Let me be explicit about this:

"In your scheme you are asking readers to intentionally misread the IPA.... it should only be used for its proper purpose of precisely representing speech. " AFAIK, the IPA has been used for diaphonemic/poly-dialectal purposes for quite some time. I gave you a historically early example. A more recent example includes Jack Windsor Lewis's A Concise Pronouncing Dictionary of British and American English (1972). Examples in other languages and between similar dialects of English are even more common.— Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]}17:37, 17 May 2010 (UTC)

I cannot get immediate access to books such as those you refer to. Are there any online sources using the IPA in the way that you describe? Martin Hogbin (talk) 21:25, 17 May 2010 (UTC)

Unfortunately, I don't know of any online availability of these resources. My local university has Lewis’s pronunciation dictionary, but I don’t know when I’ll be able to check it out. I have read a review of it on JSTOR (and a scathing one at that) that clearly states that it’s diaphonemic between AE and BE and a 2003 article by Windsor himself (in the Journal of the International Phonetic Association clearly states that the transcription system used IPA. I mentioned it as an example of the IPA being used for diaphonemic transcription. I don't know if you have access to JIPA online, but the article by Lewis is "IPA vowel symbols for British English in dictionaries" volume 33, issue 2, pages 143-152. The 1975 review was by Wolf Friederich and appeared in Foundations of Language, vol. 12, issue 3, p. 423. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 22:20, 17 May 2010 (UTC)

If there is already an IPA diaphonemic transcription scheme in existence, why do we not use that here. Martin Hogbin (talk) 22:52, 17 May 2010 (UTC)

It's my understanding that Wikipedia editors came up with what we now call a diaphonemic transcription before anyone thought to connect it to the term diaphonemic/diaphonic. For a while, we were calling it "poly-dialectal" or "accent neutral." On top of this, it seems that editors have not (yet, anyway) researched previous diaphonic transcription systems. The one done by Lewis was criticized for reasons I'm not sure about and I don't know if he even encodes all the contrasts that we do. I'm in the process of doing research to beef up our article diaphoneme, which is why I even know this in the first place. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 01:37, 18 May 2010 (UTC)

"It is not even clear how the readers are expected to get to the correct pronunciation. The IPA for English gives little clue on how to do this at all." I'm assuming you're saying it's not clear to readers, as the procedure has been explained to you here already. In "Understanding the key" (at WP:IPA for English), it says, "If, for example, you pronounce cot /ˈkɒt/ and caught /ˈkɔːt/ the same, you can simply ignore the difference between the symbols /ɒ/ and /ɔː/, just as you ignore the distinction between the written vowels o and au when pronouncing them." There are further examples in regards to a few consonants, as well as contrasts that readers may make but that we don't encode for. This is clear enough to me that it almost seems as though you haven't read "understanding the key." You don't strike me as the kind of person who would miss that sort of thing, so how could this be worded differently to be more clear to lay-readers? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]}17:37, 17 May 2010 (UTC)

It is not clear what process the reader is expected to go through to get the correct pronunciation. Do they: match the IPA symbols used to the lexical set in the key, then read the lexical set to get the sound, in which case the IPA serves no real purpose, any symbol set would do; or is there some other process such as those I described earlier, that they go through, in which case what? To make clearer what I mean, suppose the reader sees /ˈbɔːt/. What is the exact process that they go through to get the pronunciation in their own accent? Martin Hogbin (talk) 21:25, 17 May 2010 (UTC)

I believe you understand the process asked of readers as you have described it clearly and succinctly. You are correct that any symbol set would do (as I mention below with my "/&r/ and /%r/" example below. The IPA is used for a number of reasons, including those listed at the top of this talk page. Even people who disagree with the diaphonemic transcription agree that the IPA is the best choice. For readers who are more familiar with the IPA, we also have IPA chart for English dialects. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 22:20, 17 May 2010 (UTC)

I understand the process you expect now that you have made clear which of the options that proposed is the expected one. Martin Hogbin (talk) 22:52, 17 May 2010 (UTC)

"some readers may not even know how their own accent is described." Yes, and this is why we use lexical sets so that people can figure out that, since they pronounce merry and marry the same, that /ær/ and /ɛr/ indicate the same pronunciation in their own speech. This requires no linguistic training nor knowledge of the IPA. If we had a pronunciation scheme where these vowel sets were /&r/ and /%r/ it would be just as clear (though, perhaps, harder to remember). Since readers know how they pronounce common English words, using lexical sets to explain these contrasts means we don't have to know the exact accent of an individual reader in our scheme. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]}17:37, 17 May 2010 (UTC)

Again, it seems the IPA plays no significant part in this process. It is the lexical sets that do the work. Martin Hogbin (talk) 21:25, 17 May 2010 (UTC)

"I can assure you that this will go down much better than (apparently) telling people from Worcester how to pronounce the name of their home town." Are you saying that the diaphonemic transcription gives this impression, that the IPA does this, or what?

It looks that way from the reaction at Worcester. Many people will see the IPA and assume that it represents the exact pronunciation they should use, just as I would do on seeing an IPA pronunciation in my dictionary. Martin Hogbin (talk) 21:25, 17 May 2010 (UTC)

I can understand that someone might have that impression. This is why transcriptions link to here and why we want to explain clearly to such people, with as little fuss as possible, how the transcription is designed. I happen to be of the belief that, as people who become upset over something in our transcription would be likely to click the link and read “understanding the key,” we have a good setup to help people understand the way the IPA is used for transcribing English pronunciation. Is there something else we can do to make it more likely for people to get an understanding of the system we use? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 22:20, 17 May 2010 (UTC)

Martin, can you tell me what dictionary you are referring to? I have never seen a dictionary that gives a precise phonetic rendering of the pronunciation of every headword, because such a dictionary could only ever cater to speakers with one narrow accent. Instead, all dictionaries I have seen, pronouncing or otherwise, have used a broad phonemic interpretation. Grover cleveland (talk) 01:48, 18 May 2010 (UTC)

Collins English Dictionary gives IPA pronunciation for headwords. It refers to standard Southern English in the guide at the front. Martin Hogbin (talk) 22:29, 18 May 2010 (UTC)

Right now, at Wikipedia talk:Pronunciation respelling key‎, there's a discussion going on about changes to that system. Since you've brought up a policy about a respelling scheme, I’m sure you could provide some valuable insight from there as well. For many English sounds, you are correct that people automatically render spellings into their own accent (though Angr has pointed out some flaws in this, which makes respelling also imperfect). — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]}17:37, 17 May 2010 (UTC)

Thanks, I will have a look in there. Martin Hogbin (talk) 21:25, 17 May 2010 (UTC)

Any respelling system will run into difficulties because of homographs. I've simply deleted some that I've run across, because I could think of no way of fixing them apart from changing the system, which would merely shift the problem onto other words. The IPA has the advantage of not being confused with English orthography.

The debate at Worcester, BTW, is primarily from an editor who understands the system but is philosophically opposed to using anything but national transcriptions. — kwami (talk) 01:18, 18 May 2010 (UTC)

Kwami, this discussion would run much more smoothly if you and Kudpung would stop sniping at each other, and if you would stop trying to misrepresent a complicated question as a simple disagreement between two editors. The debate at Worcester has involved several editors trying to say 'this doesn't make sense', but who don't really know a whole lot about linguistics. I'm one of them - I've largely given up because the debate oscillates between 'this is really simple' and a whole load of technical jargon I don't want to untangle. I can spend my time more productively, with less frustration, elsewhere. However, I'm still watching the debate. Martin - everything you've said so far makes a lot of sense to me. I look forward to a solution coming out of this discussion. GyroMagician (talk) 07:34, 18 May 2010 (UTC)

Excuse me? That was not sniping. There was nothing malicious or disingenuous about it, simply a statement that Kudpung wants transcriptions based on national standards. AFAIK, that's a correct statement. Worcester was mentioned as an example of people being confused by the IPA conventions. However, the criticism at Worcester is almost entirely by Kudpung, and he is not confused by the system, simply opposed to it. Again, AFAIK, that's a fair statement. If you were confused by the system itself, I'm sorry, but your voice got lost in all the drama. I've found it quite difficult to understand what the point of that debate is, and no-one but Kudpung has come to me directly, so my primary impression is of him. If you are still at a loss, I'd be happy to try to explain it, to the best of my ability. — kwami (talk) 08:50, 18 May 2010 (UTC)

Whilst I ought to feel flattered that my name is being taken in vain (again), review of the pages and pages of various discussion on the fallibility of the system contrived exclusively for use in the Wikipedaia, will reveal that alas it is not Kudpung, sondern he is speaking for several editors and visitors to the encyclopedia who requested his intervention. I also feel that I have mentioned this, time and time again.--Kudpung (talk) 15:47, 18 May 2010 (UTC)

Yes, and have never explained yourself, despite numerous requests. "I don't have to explain what I want, because it's what other nameless people want, and I'm just their voice, but you have to do it anyway." That's why, despite your promising start, I no longer take you seriously. — kwami (talk) 20:52, 1 June 2010 (UTC)

Why not have the IPA just showing how it is pronounced locally (and nationally if it is pronounced differently) in articles about places? e.g. Visakhapatnam would be pronounced only one way. Maybe just respelling can be diaphonemic where required (unless it's not supposed to be?). Isn't the point of IPA that it is pronounced the same by everyone who reads it? --Joshua Issac (talk) 15:26, 9 June 2010 (UTC)

The IPA is a tool that has multiple functions. One of them is providing a more abstract transcription that accomodates multiple dialects.

The problem with giving only a local pronunciation is that locals aren't the only ones who use the name of their locality. Giving their pronunciation may omit phonemic contrasts that speakers in other areas make. For example, Dublin is pronounced by locals without a /ʌ/ since Irish speakers don't distinguish /ʌ/ from /ʊ/ (the vowel of strut and the vowel of foot, respectively). However, dialects that do make this distinction, invariably (AFAIK) pronounce it with /ʌ/. If we gave the local pronunciation and not the diaphonemic one, people would come away with an incorrect pronunciation. Giving both the diaphonemic and the local isn't a problem as long as we mark them as such.

BTW, I just went to the Dublin and see that we'll have to fix that article's use of the template a bit.— Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:49, 9 June 2010 (UTC)

Joshua, that was exactly my point, the IPA has been formulated toallow an unambiguous pronunciation. I now have a copy of the IPA handbook and note that it does indeed suggest that the IPA can be used in broader transcription schemes but there is no evidence they had anticipated its diaphonemic use over the full range of English accents. One other point worthy of note is that the handbook suggests that for broader transcriptions use of just Latin characters is preferred, this brings it more in to line with a respelling scheme. Is there room for an innovative compromise here? Martin Hogbin (talk) 17:10, 9 June 2010 (UTC)

Some more thoughts

Can one of the supporters of this scheme please answer the following.

Previous diaphonemic IPA schemes

Are there any existing diaphonemic IPA English transcription schemes? If so then why do we not base our system on them, if not then we are moving the IPA into new territory. Martin Hogbin (talk) 12:39, 18 May 2010 (UTC

I was under the impression that I answered this question above. There are diaphonemic transcriptions elswewhere, the only one I know of (after a very brief search) that encodes both American and British pronunciation is A Concise Pronouncing Dictionary of British and American English. The system we have now was, to the editors who constructed it, an exercise of WP:IAR; specifically WP:SYNTH to satisfy WP:NPOV concerns. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:20, 18 May 2010 (UTC)

Not to my knowledge. Even Wells suggests that panlectal solutions are a linguists construct and that the diphonemic approach runs into difficulties as soon as we have to deal with systemic differences between accents rather than mere differences in phonetic realisation. So it's hard to understand why the creators of the Wikipedia construct persist in maintaining that their system is a panacea. The more one reads, even as a linguist, on the IPA in this encyclopedia, the more one reaches the conclusion that the Wikipedia IPA authors are writing for themselves, (or at least for each other), rather than for the common good of the average encyclopedia visitor. This is therefore suggestive of original research i.e. 'new territory' which strictly speaking is not allowed in the Wikipedia.--Kudpung (talk) 16:17, 18 May 2010 (UTC)

With all due respect, I would suggest that the exercise in WP:IAR is really WP:OR.--Kudpung (talk) 16:38, 18 May 2010 (UTC)

Æµ§œš¹, I am not terribly worried about WP policies at this stage, but if there is an existing published scheme, why did we not just use that? Martin Hogbin (talk) 16:53, 18 May 2010 (UTC)

Like I said above, I don't think people knew about them. I'm just finding out about them myself.

Kudpung, Wells is not the only one who speaks about diaphonemic transcription. In the 1950s there was sharp disagreement about this very issue (whether a phonemic analysis can be applied to multiple dialects). I'm not sure if that disagreement has subsided or which way it's gone since. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 17:04, 18 May 2010 (UTC)

Æµ§œš¹, so the position is that, at the time this project was started, we thought were taking the IPA into new territory but we have since found out that others have been there? Martin Hogbin (talk) 17:24, 18 May 2010 (UTC)

If by "project" (which is a term that can also be used for Wikipedia itself) you mean the diaphonemic/poly-dialectal transcription, then that sounds about right. Woodstone's comment (below) also gives the impression that editors who designed this scheme felt they were only going a little bit further than what dictionaries already encode for. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:37, 18 May 2010 (UTC)

Quite right Aeusoes, and that's why I don't think we at Wikipedia should be trying to reinvent the wheel and constructing our own solution. BTW, Wells (1982) is far more modern than the 1950s, and ASFAIK he didn't evoke any earlier arguments.--Kudpung (talk) 17:24, 18 May 2010 (UTC)

I don't think a multi dialect representation is "new territory". Applying IPA is always an abstraction of the many different ways a language is pronounced by different persons and groups. It is just a matter of how wide you draw the boundary. The variants covered by the current key are not essentially more diverse than what is done in dictionaries. Listing all individual variations is just not practical. −Woodstone (talk) 18:20, 18 May 2010 (UTC)

Which dictionaries are you referring to, presumably not A Concise Pronouncing Dictionary of British and American English? Martin Hogbin (talk) 18:28, 18 May 2010 (UTC)

I'm not yet familiar with more recent debate on this issue. The tables at Pronunciation respelling for English and Help:IPA conventions for English (the latter of which is incompletely done) show the contrasts made in various dictionaries compared to our own.

Does anyone know if the similarity of dictionary.com's IPA system is copied from us or not? They're remarkably similar. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:37, 18 May 2010 (UTC)

It was already there in dictionary.com at the time we started using it in WP. −Woodstone (talk) 20:26, 18 May 2010 (UTC)

How do you get the IPA pronunciation in dictionary.com, I can only see respelling? Martin Hogbin (talk) 21:27, 18 May 2010 (UTC)

There should be clickable text that says "show IPA" in brackets next to the pronunciation. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 03:10, 19 May 2010 (UTC)

And there is a question mark bringing you to the key. −Woodstone (talk) 07:29, 19 May 2010 (UTC)

I am not convinced that there is anything diaphonemic about the scheme used by dictionary.com, it uses the IPA with a key referring to standard lexical sets but this does not prove any intention to cover more than one accent. Collins dictionary does just the same thing but refers only to Southern English pronunciations. Martin Hogbin (talk) 18:03, 19 May 2010 (UTC)

If we are to understand a "diaphonemic" system to be one that relates multiple dialects directly to a single transcription on the basis of which they may be compared and derivable into their distinctive phonological characteristics, then all dictionaries do this to some extent. If we are to take "diaphonemic" to mean that all contrasts from all dialects are represented, then not even our own system is such. We can place pronunciation schemes on a gradient between very specific to one dialect and incorporating all dialects. Help:IPA conventions for English shows that we only go a little bit further than dictionary.com and the symbols we use for each diaphoneme aren't all that different from those used at dictionary.com. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:41, 19 May 2010 (UTC)

Ausoes, you say that you are not terribly worried about policy at this stage. Ironically, we have been told repeatedly throughout this talk page that the contrived sytem is policy, is laid down in the MOS, and therefore cannot be changed. I think what a lot of us are asking is for som diffs to wherever such a policy was made my a consensus. Some contrubitors to this discussion have suggested that your consensus is nothing more than an unchalleneged decision by two or maybe three editors who game the system by saying that no comment means tacit consensus.--Kudpung (talk) 00:41, 20 May 2010 (UTC)

Kudpung, no-one has said it can't be changed. Policies are changed all the time. But to do so you need consensus. As for "unchallenged decisions", that's practically the definition of a consensus! You need to approach this as anything else: change through convincing the community, not change because in your opinion you're right and they're wrong. — kwami (talk) 00:56, 20 May 2010 (UTC)

Kwami, I'll ask you yet again to discuss the topic and/or related policy, and not the editors in these pages. At the root of the problem is that some editors game the system of Wikipedia's decision-making mechanism; any clique of two editors appear to be able to write a rule, vote it among themselves, and declare it "consensus". I want to see the diffs - It appears to several editors that that is what happened with the adoption of the policy concerning the use of the IPA in Wikipedia articles and the development of its key. Morover, most of the work in this encyclopedia on the IPA is the effort of a single contributor and containsa lot of as yet, unchallenged POV and/or OR. We all hope that this current debate will be more open, that the criticisms and comments of all participants will be heard, and that the adoption of ANY new schemes will be met by a clear acceptance of a representative number of editors, authors, and encyclopedia visitors. By contrast, the recent adoption of a new BLP policy was the work of over 400 contributors and commentators - that's what I call collaboration and consensus.--Kudpung (talk) 03:48, 20 May 2010 (UTC)

My god, what hypocrisy. To ask us not to discuss other editors, and then to launch immediately into discussing other editors. And of course with your usual gross misrepresentations. If you don't want it to be personal, quit making it personal. — kwami (talk) 04:10, 20 May 2010 (UTC)

Minor correction: it was Martin (telling me) who wasn't concerned about policy. I believe it was in the sense that he was concerned about feasability first.

Kwami is right to point out how consensus works. You say that the current system is propped up by 2 or 3 editors who act in bad faith. I'd like you to count the editors who participated in the policy's proposal. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 04:37, 20 May 2010 (UTC)

Thanks for that link. It shows that many editors were against the system from the start. Martin Hogbin (talk) 14:31, 21 May 2010 (UTC)

Yes, many editors were opposed to it in the beginning. You may notice that Angr was one of them. You may want to ask him what changed his mind. There's another editor who was also opposed to the diaphonemic system when he first discovered it but has come around. I'll give you a hint, his name also starts with A. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:49, 21 May 2010 (UTC)

Lexical sets

Do you all agree that the use of standard lexical sets is the best way to enable a reader to obtain a pronunciation in the accent of their own choice?

If that is agreed then this discussion becomes one of deciding the best way to encode that information in the articles.Martin Hogbin (talk) 12:39, 18 May 2010 (UTC

One of the things we're working on implementing is {{IPAc-en}}, which encodes lexical sets in mouseovers. Thus, instead of /ʃɪər/, for shear, we have /ʃɪər/. You can then hover the cursor over the transcription to see what the characters mean. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:20, 18 May 2010 (UTC)

That is a neat idea, but why bother with the IPA, why not just hover the mouse over the original text?

Do I take it then that you agree that the use of lexical sets is the best way to enable a reader to obtain a pronunciation in the accent of their own choice? Does anyone disagree? Martin Hogbin (talk)16:44, 18 May 2010 (UTC)

I suppose it's because the diaphonemic transcription came after we'd already been using IPA for so long. I do agree that lexical sets are the best way to enable a reader unfamiliar with the IPA to do this. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 17:04, 18 May 2010 (UTC)

I do not agree. I have suggested simply showing both phonetic transcriptions where the pronunciation in AE and BE is significantly different; and dropping the idea that a final /r/ should be included when it is not pronounced. There is no reason why an average dictionary or encyclopedia user should have to learn all about linguistic jargon and the use of yet another construct to be able to get a fairly accurate idea of the pronunciation. This seems to be the approach favoured by the majority of readers who have ventured comment in this and other related discussions. However, if one wants to get really technical, we could introduce a system where the site software recognises from the IPA, the language zone, and displays the appropriate IPA transcription. --Kudpung (talk) 16:56, 18 May 2010 (UTC)

Æµ§œš¹, maybe you have misunderstood my question. I was simply asking if you thought ultimate reference to a (Latin) lexical set was the best way to enable readers to get a pronunciation in their own accent. Martin Hogbin (talk) 17:20, 18 May 2010 (UTC)

Yes, I agree that it's the best way. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:37, 18 May 2010 (UTC)

So the question now becomes, 'What is the best way to direct the reader to the set of lexical sets that will define the pronunciation?'.—Preceding unsigned comment added by Martin Hogbin (talk • contribs)

Mouse-overs with lexical sets are great when they work, but aren't always practical. I have great difficulty using them with my current setup; they're basically too much of a pain to bother with. A footnote would be much easier. But the argument for abolishing the IPA strikes me as a bit like saying that since the majority of readers are unfamiliar with the metric system, we should abolish it and replace it with a comparison to standard measures. Is it too much to ask readers to be educated enough to use either the IPA or the metric system, especially when we provide them with links to a key that explains them? — kwami (talk) 23:01, 18 May 2010 (UTC)

The point is that you are changing all the current IPA transcriptions to your own version of the IPA. My question was asking how readers are expected to know how to pronounce your particular version of the IPA. The answer given to me was by reference to the standard (Latin) lexical sets in the 'IPA for English' key. Do you agree with that? Martin Hogbin (talk) 23:16, 18 May 2010 (UTC)

Also, as far as the "r which isn't pronounced" (notice again that no-one is objecting to the h which isn't pronounced!), that will be a problem regardless. Look at the argument at Worcester: the British Library and several dictionaries would place an /r/ there. Other dictionaries would drop it, but I don't know if they claim that's phonemically accurate. So even if we transcribe Worcester in RP, should it have a final /r/? — kwami (talk) 23:01, 18 May 2010 (UTC)

I was not asking about any specific case. The question is, 'How does a reader know how to pronounce your IPA transcription in their own accent?'. What is your answer? Martin Hogbin (talk) 23:16, 18 May 2010 (UTC)

Using the new template {{IPAc-en}}, the reader would see for example /ˌæləˈbæmə/, and by mouse-over would get an example word for every phonemic segment. Because the explanation of the symbols is based on real words (a lexical set if you like), the adaptation to the reader's dialect would be automatic. −Woodstone (talk) 07:56, 19 May 2010 (UTC)

Yes, that is what I already understand but I just wanted to make sure that that was kwami's understanding too. Perhaps kwami would confirm this? Martin Hogbin (talk) 08:55, 19 May 2010 (UTC)

That and this key, which does the same thing. You always need definitions with a phonemic transcription. — kwami (talk) 09:34, 19 May 2010 (UTC)

Technical solutions

Has anyone looked at possible technical solutions, such as a 'preferred accent' user setting so that articles would only show the IPA pronunciation in the accent of their choice? Martin Hogbin (talk) 12:39, 18 May 2010 (UTC)

According to a completely unsourced (needs PRODing BTW) article in the Wikipdia: A diaphonemic transcription... would thus cover both (and other) dialects, but exactly describe neither. Diaphonemic transcription is useful for dictionaries which are not intended for a particular dialect. In actual fact therefore, it is totally useless for the dictionary user. Most quality dictionaries are edited either for a specific language region, such as AE or BE, or, where differences in pronunciation are important, they show both. I have suggested that we show both. It wold not pose any particularly technical problems.--Kudpung (talk) 16:34, 18 May 2010 (UTC)

But it would pose problems, because by what measure do you limit it only to AE (General American?) and BE (Received Pronunciation?), and not the myriad other varieties? — ˈzɪzɨvə (talk) 16:53, 18 May 2010 (UTC)

I don't think it would. Experience made by publishers of leading dictionaries, and isogloss research has shown that by and large, BE and GAE account for the vast majority of use worldwide, and in fairly equal proportions. I think it would be fair to assume that we at Wikipedia can go along with what the professional dictionary publishers are doing.--Kudpung (talk) 17:00, 18 May 2010 (UTC)

On top of that, there is the appeal to verifiability. AE and BE are the accents most widely available in dictionaries. It's also fairly easy to encode for several other varieties once you accomodate those two. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]}

Surely we could have a table of several pronunciations but the reader would only see one according to their set preference? Martin Hogbin (talk) 17:21, 18 May 2010 (UTC)

(repeat of an earlier post that got missed) I have suggested simply showing both phonetic transcriptions where the pronunciation in AE and BE is significantly different; and dropping the idea that a final /r/ should be included when it is not pronounced. There is no reason why an average dictionary or encyclopedia user should have to learn all about linguistic jargon and the use of yet another construct to be able to get a fairly accurate idea of the pronunciation. This seems to be the approach favoured by the majority of readers who have ventured comment in this and other related discussions. However, if one wants to get really technical, we could introduce a system where the site software recognises from the IPA, the language zone, and displays the appropriate IPA transcription. --Kudpung (talk) 17:29, 18 May 2010 (UTC)

Do you mean when /r/ is not pronounced by speakers of AE or by speakers of BE? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:44, 18 May 2010 (UTC)

I think that Kudpung's idea, a system whereby the appropriate IPA for the user's own dialect is shown, is an excellent one. Ideally this could be implemented through user preferences. I'll take a look at the technical possibilities, although I'm no Mediawiki technical expert. Grover cleveland (talk) 20:17, 18 May 2010 (UTC)

You do realize that the vast majority of our readers are not logged-in editors and therefore have no Preferences to set? +An gr 20:42, 18 May 2010 (UTC)

That could be a problem. Martin Hogbin (talk) 21:29, 18 May 2010 (UTC)

Two problems I see, besides user prefs:

The pronunciations on WP are quite erratic. We're unlikely to get three transcriptions for each. What started off the transition was Australian editors insisting that their dialect be represented too. They used a transcription system that was nearly unintelligible to those of us only versed in RP and GA. And if we have three transcriptions for three nations, why not more? That could quickly become a real headache.
We can't predict whether a reader would like the local pronunciation or their own dialect. I expect that often the would want both. Although perhaps a bit extreme, someone once asked me about Montreal, and I had no idea what they were talking about, because I didn't recognize the pronunciation. I'm glad now that I know it. I would never use it, as it would sound pretentious, but I could see similar things happening with place names in other countries. So I'd want both in an entry.

— kwami (talk) 22:42, 18 May 2010 (UTC)

is not really a problem. If Australians want the Australian pronunciations to be present they (or someone else) have to add them. Similarly Cornish, Cockneys etc.
I would suggest giving the local pronunciation in addition to the selected one. Martin Hogbin (talk) 22:54, 18 May 2010 (UTC)

Okay for #2. As for #1, how would that work, then? I don't think we'd all want to see a paragraph of the pronunciation in various dialects. If we only display one, then we can either go with user prefs (I expect the default might need to be General American), or go by IP address. I suppose then that we could have it so that readers in Australia would see the Australian IPA if that's available, the RP IPA if not, and the GA IPA if the RP is not available. They'd all have to be labeled, so for one word you'd get "Australian English pronunciation X", for another "Received Pronunciation Y", and for yet another "General American Z". Or maybe we could set it up so that the diaphonemic transcription is displayed unless the local one is available. Either way, the reader will need to learn several different IPA transcription systems rather than just one. And for all I know, English readers might still end up with an /r/ at the end of names like Worcester, depending on which dictionaries and institutions we take to be authoritative. — kwami (talk) 23:10, 18 May 2010 (UTC)

User prefs would be the way to do it. If somebody sets their prefs to Cockney and finds that the most of the pronunciations are missing that is their hard luck. They can either lump it or start adding them or change their setting to RP.

Really, my point at this stage is simply that we should be discussing other options before you force your scheme on the whole of WP. Martin Hogbin (talk) 23:29, 18 May 2010 (UTC)

You're a couple years too late for that. This is the established consensus. Your point can only be to change to a new consensus. Which is fine, consensus does change, but in the meantime I'm going to continue to assume the validity of our established conventions. — kwami (talk) 00:00, 19 May 2010 (UTC)

There are no established conventions. There is a consensus on this page by a small number of regular editors to use this system but the record shows a continuous stream of dissenting editors even on this page. There have been a number of heated discussions and edit wars on article pages. I suggest that much wider consensus is required before this scheme is forced on the whole of WP. Perhaps an RfC is called for. Martin Hogbin (talk) 09:00, 19 May 2010 (UTC)

Since this transcription system is present on thousands of Wikipedia pages, and silence can be construed as a form of consensus, the consensus is not isolated to just this talk page. . Kudpung mentioned something about an RfC but I haven't seen it come about yet at Wikipedia:Requests for comment/Language and linguistics. It looked like he put some work into collaborating with other editors on the proposal, too.

BTW, have you checked out the proposal and discussion of this system? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:28, 19 May 2010 (UTC)

Kwamikagami: the current IPA format is sufficient to derive pronunciations for RP and GenAm. AusE and NZE are largely (though not wholly) derivable from RP. So, in theory, the information is already there, if we can process and present it appropriately. Grover cleveland (talk) 00:17, 19 May 2010 (UTC)

I believe you're right. And the only thing lacking for extending it to GA would be the dropped ars and aitches, which is why we decided to go with an RP representation modified only to accommodate rhotic and aspirated dialects. (Oh, and to maintain horse-hoarse, since that's easy to do from our sources.) — kwami (talk) 00:55, 19 May 2010 (UTC)

Bad news. I had hoped that it would be relatively easy to write templates that, given the current IPA-en format as input, would be able to generate either RP or GA output (or both). This would mean that we wouldn't have to go back and re-edit all the current pages with pronunciation information. In Perl such a function would be easy. While it is probably not impossible here, the severe restrictions imposed on the string-processing functions currently available in Wikipedia mean that it's far harder than I had hoped :( I haven't given up, though. Grover cleveland (talk) 00:06, 19 May 2010 (UTC)

Yeah, I've looked into doing this kind of thing before, and gave up. But then, you probably know what you're doing better than I. — kwami (talk) 00:19, 19 May 2010 (UTC)

After much blood, sweat and tears, I've come up with a proof of concept: see below. —Preceding unsigned comment added by Grover cleveland (talk • contribs)

There is one solution that would work even for non registered visitors. It's a standard procedure in web design. from the visitors IP (Internet protocol number), the area where he is connecting to the INternet would be identified, and the appropriate IPAwould be shown by his browser. Not difficult at all really. I'ts a compromise, because a few (probably not Americans) would be on the move and accessing the WP from different countries. --Kudpung (talk) 16:56, 19 May 2010 (UTC)

Wow, could you try not thinking in stereotypes? What makes you think it would be "probably not Americans" who would be accessing the Internet from different countries? For that matter, what makes you think that everyone accessing the Internet from a U.S. IP address wants GenAm pronunciations and everyone accessing the Internet from a UK IP address wants RP pronunciations? And what about people in non-English speaking countries? +An gr 17:39, 19 May 2010 (UTC)

Angr, In the interets of keeping this discussion on--Kudpung (talk) 23:58, 19 May 2010 (UTC) topic, I'll answer this in detail on your talk page.

Can I just jump in here from the ancient history of this discussion? I think that an ideal solution would be to stick with the current convention, but have some kind of "pop-up" when you click on the pronunciation that offers options for displaying it in different dialects. This would all have to be done with JavaScript, but it is definitely technically feasible with the current technology. I will try to prepare a prototype/demonstration of my concept to better explain it. nohat (talk) 17:48, 19 May 2010 (UTC)

Do you know whether Wikipedia allows a solution like this to be implemented? Grover cleveland (talk) 14:44, 20 May 2010 (UTC)

So why the IPA?

It is clear that the ultimate arbiters in the current pronunciation scheme are the standard lexical sets used in the key. This the IPA symbols serve only the purpose of encoding the sounds described by the lexical sets in the key. So why use the IPA. It might be argued that using the IPA gives s rough (but contentions) guide to the pronunciation without using the key but those who are familiar to some degree with the IPA, as it is used in UK dictionaries for example, expect it to be purely phonetic and thus find its diaphonemic use confusing. On the other hand, those who are mot familiar with the IPA will just see it as gibberish. So why use the IPA?

Oddly enough, the US dictionary symbols might be better. They were never very precise, they are familiar to Americans and they, or something similar, have been used in UK dictionaries in the past. Is there a key to the symbols anywhere?Martin Hogbin (talk) 18:18, 19 May 2010 (UTC)

Please read "The IPA is gibberish and I can't read it. Why doesn't Wikipedia use a normal pronunciation key?" at the very top of this page. +An gr 18:25, 19 May 2010 (UTC)

Perhaps you should read it, then read my question again. Martin Hogbin (talk) 19:35, 19 May 2010 (UTC)

AFAIK, non-IPA symbols haven't been used in the UK dictionaries for decades. For those in the UK familiar with the IPA, the scheme we have here is very similar to what they're already used to (the Gimson system). The most significant difference is that we also encode for /r/ in the syllable coda, which might trip them up a bit, but once they understand the key, it's not difficult for them to ignore the coda /r/'s at the ends of words and before consonants. They do it all the time when reading English orthography anyway, right? Help:IPA conventions for English and Pronunciation respelling for English if you're curious about the symbols used by various dictionaries. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 18:51, 19 May 2010 (UTC)

I am still hoping that someone will explain to me what the advantage is of the IPA when we are just using it as a code to link to the lexical sets. As has already been pointed out, we could just use numbers. Why the IPA? Martin Hogbin (talk) 19:37, 19 May 2010 (UTC)

The IPA already exists, and is an international standard for representing the pronunciations of words. Lexical sets, on the other hand, are not a feasible way to represent the pronunciation of words. nohat (talk) 20:06, 19 May 2010 (UTC)

The IPA is an international standard for phonetic transcription (the clue is in the name) that may well have been extended to phonemic transcription in some cases but there is very little, if any, evidence of usage for diaphonemic trancription. The only claimed online example of diaphonemic use of the IPA (dictionary.com) turned out to show no evidence of diaphonemic intent whatsoever. I am completely unconvinced that the IPA was ever intended or has ever been used for this purpose.—Preceding unsigned comment added by Martin Hogbin (talk • contribs)

The IPA has been used for all kinds of linguistic transcription—phonetic, phonemic, morpho-phonetic, diaphonemic, etc.—for as long as the concept for that kind of transcription has existed. No other alphabet has been used in such a widespread way for such a wide variety of transcription types. nohat (talk) 21:29, 19 May 2010 (UTC)

Exhaustive searches don't end with the internet. I've given you a couple of bound examples and, if you'd like, I can provide more examples of diaphonemic sytems of English or other languages. I also disagree that dictionary.com isn't diaphonemic. Have you seen the table at Help:IPA conventions for English? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 23:39, 19 May 2010 (UTC)

Yes I have seen that, and I do not cliam that the problem is easy but I still dispute that the IPA is the right character set to use. Clearly the IPA was originally developed to be strictly phonetic. Why else would diphthongs (usually) be represented by a pair of symbols. I would expect a scheme desigend to be phonemic to use one symbol per phoneme. A diaphonemic scheme should try to encompass problems with variations in rhoticity. The IPA is not good at this either. Martin Hogbin (talk) 11:07, 21 May 2010 (UTC)

Your logic breaks down very quickly when you consider how many different kinds of diphthongs (and affricates) there are; creating symbols for every possible diphthong would be much too cumbersome and would require a constantly evolving repertoire of symbols as linguists discovered new ones. It's the same with affricates, such as [tʃ], which are represented with two characters for the same reason. In a number of cases, whether affricates and diphthongs are phonemic is also dependant on one's analysis. Spanish, for example, has many diphthongs; are they phonemes on their own or are they instances of vowels running together? Chinese, which also has triphthongs, has a similar but more complicated situation. The affricates of Catalan are arguably phonemic, but not according to everybody who studies the language.

If your argument is that the IPA shouldn't be used for anything but phonetic transcription, then there is a lot of linguistic literature that exists in defiance of this, including the Handbook of the IPA and your position would be one of WP:OR purism. But you may be confused as to the difference between phone and phoneme, in which case I recommend you take a look at those two articles. The IPA can, and dictionaries do, present phonemic transcription. While as clear a case hasn't been made that diaphonemic transcription is widespread, you simply can't question that the IPA is used for phonemic transcription if you're going to be taken seriously. As for diaphonemic transcription, I haven't taken a look at Trager and Smith's book, but you might want to take a look at it. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:45, 21 May 2010 (UTC)

It's an internationally recognized alphabet used for pronunciation and the transcription is very similar to ones many readers are already familiar with. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 20:09, 19 May 2010 (UTC)

It is indeed an internationally recognised alphabet used for pronunciation. That is the problem. You are using it simply as a coding mechanism for standard lexical sets. Readers such as myself and many others from the UK who are familiar with use of the IPA for phonetic transcription find its use for diaphonemic trancription confusing and annoying. This is exactly what has happened at Worcester. —Preceding unsigned comment added by Martin Hogbin (talk • contribs)

Yes, that's what any transcription system would be. Including a less diaphonemic transcription system. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 23:39, 19 May 2010 (UTC)

Martin: I don't know whether this is still an active dispute. But take a look at the authoritative manual on the IPA: the Handbook of the International Phonetic Association. The very first example of its use includes a set of "Conventions" (Google Books link) which includes statements such as [t] is a voiced flap, resembling [ɾ]. This shows that the paradigmatic definitions of the IPA symbols are subject to reinterpretation in any particular usage. Grover cleveland (talk) 15:01, 20 May 2010 (UTC)

Because for most people outside the US who know how to use a dictionary, the IPA is all they need. And if you understand the IPA, it's much quicker than lexical sets. — kwami (talk) 20:13, 19 May 2010 (UTC)

This is completely incorrect if you are using the IPA for diaphonemic transcription. You have all agreed that the final arbiter of sound in your scheme is the standard lexical set. The normal values of the IPA symbols are irrelevant as they are to be overridden by the users own pronunciation of the lexical set shown on the key. This may be substantially different from the pronunciation actually indicated by the IPA. Martin Hogbin (talk) 21:10, 19 May 2010 (UTC)

I don't have The Handbook of the International Phonetic Association, which contains the official specifications for the IPA in front of me, but if my memory serves me correctly, it specifically says that there is a long history of the use of the IPA to represent phonemes and other linguistic entities that don't exactly match the official definitions of the symbols, and that that is a perfectly legitimate, reasonable, and acceptable use of the IPA. Not to mention practical and historically sound. nohat (talk) 21:34, 19 May 2010 (UTC)

Martin: consider an example. Most dictionaries that use IPA (including ones aimed at Englishmen) would represent the word "bill" as something like /bɪl/. This is despite the fact that the vast majority of residents of Southern England today, and very probably the majority of overall residents of the UK, have no phonetic consonant corresponding to /l/: their pronunciation is phonetically something like [bɪʊ] or [bɪo]. This shows that dictionaries tend to use a phonemic, not a phonetic, transcription, where in this instance /l/ in syllable coda is vocalized by rule. Grover cleveland (talk) 23:28, 19 May 2010 (UTC)

No, dictionaries use phonetic pronunciation is a specified accent. Martin Hogbin (talk) 15:39, 20 May 2010 (UTC)

No, they don't. It's widely understood that dictionaries use phonemic pronunciations in a specified accent. This is why those that use IPA present it in /slashes/. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:57, 20 May 2010 (UTC)

Martin, you asked whether lexical sets were the "best way to enable readers to get a pronunciation in their own accent" not whether they were "arbiters" (a term I'm confused by in this context). The lexical sets are a tool in teaching the system. The transcription is based largely on transcriptions already available in dictionaries like that at dictionary.com, as well as some others. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 23:50, 19 May 2010 (UTC)

I just want to say to Martin Hogbin that I think the general thrust of this section—that use of the IPA to represent linguistic transcriptions that are not purely phonetic is somehow not appropriate—is really not a supportable claim. I recommend you try a different tack in this debate. nohat (talk) 21:37, 19 May 2010 (UTC)

I hope you can excuse my being a little sceptical about the use of the IPA for diaphonemic transcription. From what is written here I might assume that diaphonemic usage is commonplace and would therefore expect there to be many online examples of such it, yet no one has yet directed me to a clearly diaphonemic use of the IPA. Perhaps you could help. Martin Hogbin (talk) 15:12, 20 May 2010 (UTC)

Martin, you deny that dictionary.com encodes a diaphonemic transcription. If it doesn't do this, then what accent would you say it encodes? — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:57, 20 May 2010 (UTC)

I would not put as strongly as 'deny' but there is no convincing evidence that the IPA on ditioary.com is intended to be used diaphonemically. They may just have chosen a mid-atlantic accent. In fact, I have found a site that uses the IPA diaphonemically but I still cannot accept that such use is widespread or even beneficial. To exixting users of the IPA it is confusing. Martin Hogbin (talk) 10:57, 21 May 2010 (UTC)

Dictionary.com marks a distinction between the vowels of moral, horse, and hoarse, as well as between word pairs like dew/do and which/witch. AFAIK, no accent does all of that. Certainly no commonly known one. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:45, 21 May 2010 (UTC)

Let's be fair first of all and admit, without thinking in stereotypes, that the needs of Americans, British, and foreign learners for a pronunciation guide are not identical. This is however not a reason to assume that the IPA is not widely used outside the US, and thus not wanted in the Wikipedia, or should be substituted by yet another contrived scheme. It is possible the majority of average American people might not have as much high frequency first-hand exposure to numerous foreign languages as, say, the Brits, the Europeans, and the Asians, where their regions are made up of clusters of small countries each with their own language and culture. In Europe for example, nobody lives more than two or three automobile hours away from another country.

As to the use of the IPA, remove the few linguists from the scene, and we will see that the vast majority of users of the IPA are language learners - and that does not make them linguists per se. Until now we have been discussing how dictionaries use the IPA , but we have not mentioned the use of the IPA in the hundreds of different language textbooks that are in use by millions of students worldwide. A look at some of them, will reveal whether the IPA use is phonetic or phonemic.

We should decide whether the focus of this discussion is a purely academic argument between linguists, or whether we are genuinely striving towards a solution that will appeal to the majority visitors to the encyclopedia, who are most probably not linguists, and taking due note of what the non linguists here in this discussion are saying.--Kudpung (talk) 00:27, 20 May 2010 (UTC)

I agree with what you have said here, Kudpung. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 04:30, 20 May 2010 (UTC)

I agree with the sentiment. I don't know about the vast majority of English speakers outside the US being language learners. I would think most are familiar with how their dictionary works. But I think it would be a fair statement of Americans. — kwami (talk) 07:30, 20 May 2010 (UTC)

Kudpung's is an excellent comment. I think all of you are trying to produce a system that is useful for non-linguists. The problem is that you are linguists, so it's hard to remember what the rest of us don't know! As a non-linguist and reluctant language-learner, I have a vague knowledge of the IPA from using a dictionary, but mainly from attempts to learn other languages (or at least a few phrases). I'd guess I'm pretty common in this - i.e. as Kudpung says, "the vast majority of users of the IPA are language learners". However, I think the problem of describing words in an unknown language is different to describing unknown words in a known language - as Kwami has repeatedly pointed out, if you say an English word to me, I'll quickly repeat it adjusted to my own accent. If you tell me a word in Polish, I'll try my best to repeat the sounds I think you made. I guess this is the crux of your phonetic/phonemic discussion? GyroMagician (talk) 11:30, 20 May 2010 (UTC)

I think that anyone who has used IPA when learning a language, remains an IPA user for life. Therefore the learners are not the majority of users. Just as native speakers, the people using a secondary language, will occasionally meet new words, for which pronunciation cannot be easily predicted from spelling. It is for those cases that in this English enclopedia, an English phonemic pronunciation is needed. Readers will usually not be interested in dialectical differences, except where they break regular patterns (like in tomato). −Woodstone (talk) 13:58, 20 May 2010 (UTC)

Kwami, I don't know whether my experience is atypical, but I will share it anyway. I grew up in the UK, where I learned several foreign languages in {secondary/high} school. I also did a lot of fairly advanced classical choral singing in multiple foreign languages. In all that time I never once heard the acronym "IPA", let alone learned or used it. I later moved to the USA and almost immediately was confronted with widespread use of IPA in classical singing in non-English languages. It was also forced to learn it to take an acting class that I signed up for. So, based on my personal experience, IPA is much more widely used in the USA (at least in singing and possibly acting) than in the UK. Of course that is just one person's anecdotal evidence, but it does lead me to question whether the USA is really so relatively benighted when it comes to knowledge of IPA. Is there any real evidence on this front, or is it possible that you have been burned by a few confrontations with some US-based Wikipedians? And is there any evidence that a different system would both be as accurate as IPA, and be so widely accepted? Given the number of articles with IPA-en transciptions, I am not surprised that the system generates queries. Grover cleveland (talk) 14:30, 20 May 2010 (UTC)

Grover, you seem to be completely missing the point here. The term 'accurate' has no meaning when applied to the IPA used diaphonemically. With the system set up as it is now, we could replace the IPA symbols with numbers and it would be just as accurate. Martin Hogbin (talk) 15:19, 20 May 2010 (UTC)

Martin, with respect you are the one missing the point. Your criticize the current Wikipedia convention for IPA in English by saying that "[w]ith the system set up as it is now, we could replace the IPA symbols with numbers and it would be just as accurate". To the extent that this critique is valid, it also applies to 99% of uses of IPA in published literature, to all uses of IPA in dictionaries, and to all uses of the IPA given as examples in the chapters of the Handbook of the International Phonetic Association dedicated to separate languages, in fact to all uses of the IPA except those known as "impressionistic" transcriptions. Please read the brief section entitled "Broad and narrow transcriptions" from the Handbook of the International Phonetic Association. You can find it on Google Books right here. In particular, note the paragraph beginning Any transcription is connected to a speech event by a set of conventions. For an example of such conventions in use, see this section from the same Handbook. Cheers. Grover cleveland (talk) 17:06, 20 May 2010 (UTC)

Martin, your objection is not coherent. The IPA is just a set of symbols. Each symbol even has an official number associated with it. The symbols also have nominal phonetic denotations, but it is understood that no use of the IPA is an exact mapping from the symbols to the nominal phonetic denotations. The nominal denotations are just a guide to help you select the symbol that is best suited for the transcription task at hand. That is, you should probably use /i/ to transcribe a high front unrounded vowel and not a voiceless velar fricative. In the case of dictionary transcriptions, they're all what we're calling here "diaphonemic", meaning that the symbols are used to designate sounds that may be different from dialect to dialect. That is why all dictionaries have phonetic transcription keys showing example words for each sound in their transcription system. But which symbols are picked is ultimately arbitrary. You have to be arbitrary if you're picking symbols that represent more than one sound. This is already the case when picking symbols for even a single dialect because as a phonemic transcription system you have to pick one phonetic representation for each phoneme, which could have multiple phonetic realizations. For example, the /l/ phoneme is usually transcribed as /l/ even though it is often produced as [ɫ] or [o] in some positions. So whether the system is phonemic or diaphonemic, sometimes the nominal phonetic denotation for symbol used does not match the actual phonetics of the sound it represents. No matter what IPA transcription scheme is used, this will always be true, so the fact that it is true of the scheme we are using now is not a valid indictment of that system. nohat (talk) 17:55, 20 May 2010 (UTC)

So, in conclusion, we use IPA because it is the international standard for phonetic transcriptions, whether those transcriptions be narrow impressionistic transcriptions or broad (dia)phonemic ones. The IPA is used for all kinds of phonetic transcriptions, regardless of the level of abstraction they represent. There is no point at which one says "this is too abstract a level of phonetic transcription for IPA", and even if there were there is no other system to use. nohat (talk) 18:00, 20 May 2010 (UTC)

That is your opinion and it is presumably the opinion of those who developed the current scheme but there are many editors who disagree with you. I have still seen no convincing evidence that IPA has be used for diphonemic transcription anywhere outside this project. Martin Hogbin (talk) 07:53, 21 May 2010 (UTC)

A concise pronouncing dictionary of British and American English (1972). This is the third time I've referred this book to you. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 16:45, 21 May 2010 (UTC)

Yes, I did notice the first two times and I will take your word for it that it is indeed diaphonemic but if (wide) diaphonemic use of the IPA is common then I am surprised that there are not more online examples. Martin Hogbin (talk) 21:51, 21 May 2010 (UTC)

All dictionary uses of IPA are diaphonemic. The only difference is which dialects are included in the diaphonemes. nohat (talk) 18:44, 21 May 2010 (UTC)

Yes, you are absolutely right. The question is, how wide a range if dialects can be described with one diaphonemic set of symbols? Would you agree that it is possible to have too wide a range of dialects to be represented by one set of IPA symbols? Martin Hogbin (talk) 21:51, 21 May 2010 (UTC)

A helpful term here might be "core inventory" (or "common core"), which refers to the phonemic contrasts shared by all the dialects represented. When a diaphonemic transcription represents dialects that share all phonemic contrasts, there's very little doubt (especially by dictionary publishers) that it's feasable. When you start getting contrasts that only some of the dialects represented make, it becomes more contentious. I don't think the answer to the line you draw on what's feasabile and what isn't is something that can be said objectively, but a search within the literature might be illuminating in this regard. — Æµ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 23:01, 21 May 2010 (UTC)

I think as long as several dialects are said to belong to the same language, I think it should be reasonable to expect that a common broad transcription should be possible. And even beyond that, closely related languages can have a common representative phoneme set. −Woodstone (talk) 12:36, 22 May 2010 (UTC)

I think the point at which it becomes unreasonable to make a diaphonemic representation is not at a certain range or breadth of dialects, but rather at a place such as when you start to contemplate making a 4-way diaphonemic contrast between the vowels of spa/pat/bath/pasta ( BrE–AmE /ɑ–ɑ/ /æ–æ/ /ɑ–æ/ /æ–ɑ/), and use a different symbol for all four. However, that's just where I draw the line, and I have certainly seen proposed diaphonemic representations that encode exactly this 4-way contrast. Probably the person who made that proposal would draw the line at a /ɛf–juː/ diaphoneme for lieutenant. The point is that the line at which diaphonemic systems are unreasonable is when you are loading your diaphonemic inventory so much that it is encoding not just phonetic differences but also lexical ones. This is where I think the diaphonemic principle becomes unreasonable. The existing system used on Wikipedia, however, is well to the reasonable side of this line. nohat (talk) 17:38, 22 May 2010 (UTC)

I am not talking about the point where the phonemically principle becomes unreasonable, I am talking about the point where it becomes unreasonable to use the IPA for diaphonemic transcription. The point is that the IPA symbols already have relatively fixed phonetic values. If the IPA is used for wide diaphonemic transcription we reach a point where, for some accents, the pronunciation represented by the standard phonetic interpretation of the IPA is very different from the diaphonemic interpretation. Martin Hogbin (talk) 19:03, 22 May 2010 (UTC)

They only have relatively fixed values when used phonetically. When used phonemically, they're quite loose. For example, in the IPA handbook itself they have /c/ for [tʃ]. Loads of texts use /r/ for English [ɹ], and /o/ for anything from [ɔː] to [əʊ]. Spanish /β ð ɣ/ are used for [b d g]. In conventional RP, /r/ is even used for silence. That's inherent to phonemic description, but also a long long tradition of the IPA itself. — kwami (talk) 19:30, 22 May 2010 (UTC)

—Digression Alert— Do Spanish-ists really use /β ð ɣ/ for [b d g], or do you mean that the other way around? Lately, I've noticed that /β ð ɣ/ would make more sense, since they're only stops in very limited circumstances, but I still thought the norm would be /b d g/.

I've seen it both ways. Then general tendency is to use ASCII letters for phonemes wherever possible, and that sometimes overrides the 'elsewhere' principle. Usually does in this case. — kwami (talk) 19:50, 22 May 2010 (UTC)

And was there ever a serious discussion of using /oː/ and /eː/ for what we currently have as /oʊ/ and /eɪ/? Many texts do transcribe them that way. — ˈzɪzɨvə (talk) 19:42, 22 May 2010 (UTC)

That was years ago, but they were at least on the list of options we chose from. I think people wanted to stick w s.t. that made it clear these were diphthongs, so readers wouldn't mistakenly think IPA [e] and [o], when used for other languages, were close to English /e:/ and /o:/. — kwami (talk) 19:54, 22 May 2010 (UTC)

OK, I will have a look at the IPA handbook. Martin Hogbin (talk) 19:34, 22 May 2010 (UTC)

AFAIR, /c/ was used that way for Hindi. But it's not uncommon with other languages. — kwami (talk) 19:50, 22 May 2010 (UTC)