Talk:International Phonetic Alphabet for English/Archive 2

Archive 1

Archive 2

Archive 3

ʊə sounds

The chart entry for ʊə (near the bottom) says that in British English it is cure, poor and jury but those sounds aren't the same at all. Cure and jury are more cyore (or cyurh) and jyurie whereas poor is the same as pore

Which sound is the correct one for the IPA symbol? From the Row below which is Australian English, I would guess that it is the poor sound --MyNameIsClare 12:33 14 Apr 2005 UTC

It's true that many British speakers use [ɔː] in cure and poor, so that poor is homophonous with pore and cure rhymes with both. The symbol [ʊə] represents the sound used by those speakers for whom poor is still distinct from pour and for whom cure rhymes with poor but not with pore. The sound spelled [ʊə] is a diphthong that begins with a vowel like that of foot, or partway between that of foot and that of goose, and then moves to the schwa sound of sofa; it sounds a bit like "oo-a". Some people in England do still pronounce poor and cure with that vowel, but it's losing ground. --Angr/_comhrá 12:59, 14 Apr 2005 (UTC)

Oh okay, I see what you mean. Yes some people do say cure and poor like that, but not many. It doesn't seem helpful to have those examples in the table without mentioning that as I think it would confuse people who pronounce them in the more common way. And I don't think I have ever heard anyone say jury like that! --MyNameIsClare 13:10, 14 Apr 2005 (UTC)

I must be another "just weird" person, but I am an American (Californian) speaker who pronounces poor [pɔɹ], cure [cjʊɹ], and jury [dʒɚi]. Anyone care to explain this away? Is this dialectal or idiosyncratic? Also, my pronunciation of the vowels in father and car is more like [ɐ]. --PlatypeanArchcow 01:35, 27 Apr 2005 (UTC)

It can't be explained away. The ugly truth is that there is virtually no word in English that consistently has /ʊr/ for rhotic and /ʊə(r)/ for nonrhotic speakers. I predict that in a century or two there won't be an accent of English that even has that phoneme anymore. --Angr/_comhrá 06:11, 27 Apr 2005 (UTC)

I'm a British/English Yorkshireman, and have spoken the language for over 50 years. Most of this article, to me, at least, reads as clear as Chinese in an Urdu dialect. The article needs copyediting and whittling down to a basic page, before it becomes an encyclopaedia in it's own right.

I know that this anon comment is from long ago, but I thought I would respond: This is a technical article about the phonetics of English. The fact that you've spoken English for 50 years doesn't necessarily make you able to read and fully understand a technical article about phonetics any more than the fact that you've lived in the physical world for 50 years should make you able to read and fully understand technical articles about physics. The page is pretty tight as it is—everything on this page is explained and can be read about more fully by following links. If it were whittled down any more it remove useful information. Nohat 05:32, 13 Jun 2005 (UTC)

May I add that the fact that you've spoken English for 50 years doesn't necessarily mean that you can appropriately use "it's" and "its".--PizzaMargherita 21:21, 22 October 2005 (UTC)

Further Discussion

AxSkov: It is probably worth saying what we wish the article to say before any further changes. As I understand it:

• In Australia, in words like JURY, CURIOUS, CURE, PURE, LURE, where there is or was a /j/, the sound is pronounced OO-r or OO-a (not a diphthong). SURE lost the /j/ early enough it was not protected from the /U@/ > /o:/ change; whereas TOUR is usually irregularly pronounced OO-a. (I speak this dialect.)
• In America, in words like SURE, JURY, CURE, the sound is pronounced UR. I don't know how this dialect treats words like LURE that lost the /j/ and didn't get a post-alveolar. This is based on what I've heard around the Internet (and the original article), so my understanding here might be wrong.

What do you dispute, and why?

Sorry, I was supposed to remove 'jury' from AuE, I'll correct my mistake. Since when has jury used /j/ in it? I'm pretty sure it has never had /j/. – AxSkov (T) 03:35, 25 Jun 2005 (UTC)

Well, it doesn't currently (by neither does 'lure'). The spelling is indicative of a former /j/, though, and yod-dropping in AuE is usually cited as having occurred after /s z l r S Z tS dZ T/ (I'll try and find an example if you really want, but it'll have to wait a bit I think). There's also the pronunciation example given higher in this section (as JYOORIE, I think). Also there's the pronunciation of 'jury' itself, as joo-ry, not jor-ry though I spose that's begging the question or something (tour's slightly different because some pronounce it tore, but no-one says jor-ry).

No, don't worry about it, it's not really that important. That yod-dropping has occurred in most English dialects, where /j/ is dropped after /l, ɹ, ʃ, ʒ, tʃ, dʒ, θ/, but not always /s, z/.

Jury probably never had /j/ in it. The /juː/ diphthong was originally /ɪʊ/ and is still retained as such in some accents. There are still accents where choose and chews are distinct as /tʃuːz/ vs. /tʃɪʊz/, you and yew as /juː/ vs. /jɪʊ/, and rood and rude as /rɪʊd/ vs. /ruːd/. In accents where /ɪʊ/ became /juː/, though, those pairs are always identical; probably there was no stage where chews was actually pronounced */tʃjuːz/, yew */jjuː/, or rude */rjuːd/. I suspect the change was actually /ɪʊ/ > /uː/ after /j, ʃ, tʃ, ʒ, dʒ, r/ and /ɪʊ/ > /juː/ elsewhere. Yod-dropping after other sounds came later. Because of the laxing/centering effect of the following /r/, jury went from /dʒɪʊrɪ/ to /dʒuːrɪ/ to /dʒʊərɪ/, the source of all modern pronunciations (except those in /ɪʊ/-retaining accents).

As for lure, as an American, I'd only say /lʊr/, but I think I've heard /lɝ/ from other Americans and certainly wouldn't bat an eyelash if I did hear it.

Maybe no Australians say jury with the NORTH/FORCE vowel, but some Brits do. A 1998 poll of British speakers reported in Wells (2000) shows 77% prefer /dʒʊəri/, 13% prefer /dʒɜːri/ (suprisingly enough; I thought that was a uniquely North American pronunciation), and 10% prefer /dʒɔːri/. Among those born since 1973, the numbers are about 55%, 25%, 20% respectively. --Angr/_{tɔk tə mi} 08:08, 25 Jun 2005 (UTC)

What Brits do has no bearing on a discussion of AuE (tho if Brits do do that, it might be worth mentioning in the article, I dunno). As for the rest, oh interesting. 'Jury' still fits into the list (along with lure, cure) of words that are pronounced with the OO (u-dashed) sound in AuE, though. It may be best to simply say that there has been a split, firm otherwise except for 'tour', which varies. Felix the Cassowary 10:55, 25 Jun 2005 (UTC)

Suggestions for improvement

As I understand it, this article was originally intended as a quick and easy guide to the IPA as it's used in English for the sake of readers who encounter phonetic transcriptions in other articles and want to know how to read them. But now, it's a nearly impenetrable mass of information about several different English accents and fine phonetic detail, that no one who doesn't already understand linguistics can understand. Therefore, I've created a proposal for a new article here, currently at User:Angr/IPA for English (proposal) and would like to get feedback from other users of this article on the idea of replacing the current text with my proposal. The issues in English phonology that are currently discussed here can, in my opinion, be profitably discussed at English phonology (for things that are transdialectal, like aspiration, the nature of affricates, and the debate whether secondary stress really exists) or in the articles on the accents in question. So, what do you all think of my proposal? --Angr/_{tɔk tə mi} 12:22, 25 July 2005 (UTC)

I think it's a big improvement, and it (mostly) avoids one of the things I dislike about the current page, the conflation of the terms "British English" and "Received Pronunciation". --JHJ 17:10, 25 July 2005 (UTC)

Yeah, that's almost as bad as the conflation of "American English" and "General American"! --Angr/_{tɔk tə mi} 17:27, 25 July 2005 (UTC)

I quite like it. Much easier to read! Thanks a lot! However, I have four small issues...

• Firstly, it looks a bit odd with the lax/checked/short vowels so near the tense/free/long ones, and I find it a bit hard to read. I would suggest either putting a bigger gap in between (e.g. by using two horizontally-aligned tables) or, better still, using something like Template:CSS_IPA_vowel_chart with left-of-dot indicating l/c/s and right-of-dot indciating t/f/l because no English included on the page uses rounding as the only distinctor.
• Regarding the /r/, it might be worth mentioning that those that do use [r] use it as the equivalent phoneme, so it hurts even less.
• For AusE diphthongs, would it be better to say 'to rounded' and 'to unrounded', rather than 'to front' and 'to central or back' (just for reasons of conciseness).
• And, I'm not sure if I've ever seen syllabic consonants transcribed for AusE, but maybe I just haven't read widely enough.

Do you mind if I make these changes? (It'll be till this evening Z+10 before I can, so you might do it yourself ... I ask more because it's under your name and I'm not sure if that's reasonable or not.)

Felix the Cassowary 03:53, 26 July 2005 (UTC)

One question: why are the mid vowels e:, o: given as diphthongs in RP, but long vowels in US, when they're phonetic diphthongs in both? Either way would be fine, but people will assume the difference is based in the dialects themselves rather than convention. kwami 04:30, 2005 July 26 (UTC)

Okay, I've added space between the short/checked vowels and the long/free vowels, and I've changed the Oz diphthongs to "to unrounded/to rounded". I don't particularly want to mention which accents do use the alveolar trill for /r/, because that will make the discussion cluttered, and it's none of the "big three" discussed here anyway. That too is a subject better discussed at English phonology. I don't know whether AusE has syllabic consonants, I just assumed it did. Felix, when you say "button" or "bottle", does your tongue ever lose contact with your alveolar ridge between the /t/ and the sonorant? Mine doesn't; it stays firmly pressed to the alveolar ridge the whole time, so the syllabic consonant transcription is right. But if Australians generally release the /t/ and then have a second alveolar closure for the /n/ or /l/, then a transcription with schwa is right. As for the FACE and GOAT vowels, it is mostly convention that has them transcribed as diphthongs in RP but monophthongs in GenAm, but it's convention with some basis in fact. Although this varies from speaker to speaker, I think the vowels do tend to be more monophthongal for GenAm speakers than for RP speakers, especially in closed syllables. Nevertheless, I've added a note saying /e/ and /o/ are often diphthongal. (The same is true of /i/ and /u/ actually, but I'm trying to keep it simple.) Any of you, feel free to make changes directly on the page. This is a wiki, that's what it's for! And if I don't like your changes I'll just revert them, so no big deal! ;-) --Angr/_{tɔk tə mi} 05:54, 26 July 2005 (UTC)

I prefer [b6d@n]/[pæd@n] (pattern) and [bOd@l]. [bOdl=] seems unlikely (indeed, difficult!). But in any case, what's the phonetic realisation matter? Do you have any minimal pairs? (I suppose that question verges on 'original research' if its used as the basis for any decisions.)

I should also point out that 'boor' and the like are not pronounced with /U@/, any more than 'shore' is pronounced with /o@/. Present cases of /U@/ can only be replaced with /ʉːə/; the split between /U@ ~ ʉːə/ and /o: < U@/ is finished and firm. (I can remember quite vividly in primary school learning about homonyms, and one example was SURE vs SHORE, included on a photocopy from one of those books teachers get with lots of handouts to give.) Also, introductory texts on Australian phonetics often shy away from providing words that have it. The best example is probably 'tour', but as a phoneme it's very definitely on life support. I'll make this change.

(Regarding /r/, I didn't mean to say which dialects did use it, just that if a dialect does use [r], it's for the same phoneme that other dialects use [ɹ] for. Or even just to get rid of the disclaimer, simply say: frequently written as /r/ in broad transcriptions of English fullstop. But while on the subject of the consonants, do any of the major three actually use /ʍ/? And is the disclaimer that ng's only used after vowels necessary, given no equivalent disclaimer on /h w j ʍ (ɹ)/, which only occur before vowels?)

Felix the Cassowary 11:34, 26 July 2005 (UTC)

One last question. What's a 'rhoticised diphthong', and do they really need to be included separately rather than as the sequence vowel + /r/? Felix the Cassowary

Actually, regarding syllabic consonants I'm obviously wrong about never having seen it, because I've read this page before. It says it's speaker and context dependent, and I'm probably just a speaker who doesn't use the syllabic consonants at least in recitation mode. Seeing as some reputable source has included the possibility, it's probably best to include them, so I'll put them back in. A disclaimer on their use is probably more appropriate for the Australian English page, which needs major work. Felix the Cassowary 11:59, 26 July 2005 (UTC)

For some non-rhotic speakers (maybe not Australians, but certainly some Brits and some Eastern New Englanders) you get minimal pairs like Saturn /ˈsætən/ vs. satin /ˈsætn̩/ (the nonrhotic Americans flap in Saturn but not in satin, as well). And for the few people who know the word nockerl (a kind of Austrian dumpling), I suppose it would make a near-minimal pair with knuckle (using GenAus vowels, /ˈnɔkəl/ vs. /ˈnɐkl̩/. As for /ʍ/, I think there are Americans whose accent can be described as GenAm who use it. Me, for instance. And the fact that there are people in both Britain and America who hypercorrectly use it in words where it doesn't belong shows that it still carries some prestige, so we shouldn't just ignore it. Personally, I hate treating it as a labiovelar fricative anyway; I think of it as a cluster /hw/. But I included it because I'm sure if it's left out now, someone else will just add it later anyway. I only included the note about /ŋ/ because all other consonants are illustrated in word-initial position and I wanted to explain why that one isn't. We can delete if you want. --Angr/_{tɔk tə mi} 12:05, 26 July 2005 (UTC)

On syllabic consonants, my intuition (which may not say anything about the actual phonetics) is that there is an /ə/: /ˈbʌtən/, /ˈbɒtəl/, etc. But I'm not exactly an RP speaker, and the OED online 3rd edition has things like pattern /ˈpatn/, nickel /ˈnɪkl/, etc. (in their notation - they don't mark the /n/ and /l/ as syllabic). I'm not convinced by your first minimal pair, though, at least as far as RP (and indeed the rest of non-rhotic BrE) is concerned: surely Saturn rhymes with pattern, and satin is [ˈsatɪn] for me.--JHJ 16:58, 26 July 2005 (UTC)

Well, Australian English has removed some other traces of English's rhotic past that I understand RP and nonrhotic American haven't, namely if a word ends in vowel (other than /æO/, /ʉː/, /i:/, /æi/, /Ae/), then it gets a linking R no matter what: 'saw it off' is [so:-r\-@t-Of]. So probably phonemic syllabic consonants are either almost gone or completely gone, and any variation is just allophonic for Australian English. ’Twould make sense, if I say it like that (to me at least ;) Thanks for your clarifications & patience. Felix the Cassowary 03:30, 27 July 2005 (UTC)

Intrusive R is hardly unique to Australian English; I'd venture a guess that the vast majority of English people and Eastern New Englanders also put an R in phrases like "saw it off". --Angr/_{tɔk tə mi} 05:18, 27 July 2005 (UTC)

Didn't mean to say it was (I know it's used in various places in England), however I have been assured by possibly wrong people that it is considered wrong/not used in RP and not used by non-rhotic Americans. My sources were just people though, who could be wrong, describing speech of other people. (Most of them were quite convinced of it, and not living in Eastern New England it's not really possible for me to make my own observations.) Tried to draw a comparison, wasn't claiming it was true. Felix the Cassowary 07:35, 27 July 2005 (UTC)

Nonrhotic U.S. Southerners don't have intrusive R (or linking R either), but nonrhotic New Englanders do. John McCarthy, a native of eastern Massachusetts, has written extensively on intrusive R in his dialect. Wells says native speakers of RP tend to use intrusive R, but adoptive speakers of RP tend not to. --Angr/_{tɔk tə mi} 08:38, 27 July 2005 (UTC)

Ah, thanks, shows what trusting random people on the net can do :) Obviously I'll have to find things by McCarthy and Wells too, else I'll just be randomly trusting a different person :)

See J. C. Wells Accents of English vol. 2 (Cambridge University Press, 1982; ISBN 0-521-28540-2), pp. 283-85. McCarthy's papers are: "Synchronic rule inversion" (Berkeley Linguistics Scoiety 7 (1991), 192-207), "A case of surface constraint violation" (Canadian Journal of Linguistics 38 (1993), 169-195), and "A note on Boston r and the elsewhere condition" (1999), hopefully still available online at http://www-unix.oit.umass.edu/~jjmccart/appendix.pdf.

There's also an article by Wells online at http://www.phon.ucl.ac.uk/home/wells/rphappened.htm, which talks about intrusive r, amongst other things.--JHJ 12:09, 27 July 2005 (UTC)

As for rhotacized diphthongs, I grant that term isn't used often, but there is phonetic evidence that the NEAR, SQUARE, CURE, NORTH/FORCE, and START vowels are really diphthongs endings with a nonsyllabic [ɚ] in American English. Of course, nonsyllabic [ɚ] is not phonemically distinct from /ɹ/, but the phonetics of the American r-sound are quite different in onset and coda position. (Somewhere on the web there's a sound file of John McCarthy saying "paint the loofa red" and "paint the loo for Ed" to illustrated the difference.) John Kenyon in his book on American pronunciation describes the vowels in question as diphthongs, and I have a publication (available here; let me know if you need a password to access it) where I argue that /ɑɹ ɛɹ ɔɹ/ are entirely contained in the syllable nucleus, rather than having the vowel in the nucleus and the ɹ in the coda. The evidence isn't so clear for /ɪɹ ʊɹ/ but for purposes of this page it's easiest to group them all together. --Angr/_{tɔk tə mi} 12:26, 26 July 2005 (UTC)

[ʍ] is an approximant, not a fricative. It's only called a fricative out of tradition, because for quite a while most fricatives and approximants were not clearly distinguished in the IPA (as is still the case when they're dorsal or radical and voiced).

It's always seemed to me that /ɚ/ is a diphthong like /i/ and /u/ in my dialect. I get a similar offglide with pearl to what I get with peel, for instance. With purr it's difficult to feel, but isn't far from that of pea. Perhaps in a close transcription it would be [ɚɹ]? kwami 12:48, 2005 July 26 (UTC)

What's happened to these suggestions for improvement? Are they going to moved over soon? — Felix the Cassowary 05:36, 10 August 2005 (UTC)

I intended to but then I was without internet access for a month. I would still like to replace this page with what's at User:Angr/IPA for English (proposal) if no one objects. --Angr/_{tɔk tə mi} 18:38, 1 September 2005 (UTC)

the two [aj] sounds

Angr reverted the inclusion of the 2nd [aj] phone with the statement "[@I] is an allophone not a phoneme". However, this is not entirely true. In most US English ("most" in the geographical sense, anyway), these are distinct phonemes, although only marginally so. Like the velar vs. palatal fricatives in German, there are few minimal pairs. One is

rider [ˈɹʷɑjɾ.ɹ̩]

writer [ˈɹʷɐjɾ.ɹ̩]

For people who flap their alveolars stops, the distinction is entirely in the vowel. Even if you chose to transcribe the distinction as one of length, there are still two [aj] phonemes. kwami 18:37, 2005 July 25 (UTC)

The surface distinction may be entirely in the vowel, but the phonemic distinction is entirely in the consonant: rider /ɹaɪdəɹ/ vs. writer /ɹaɪtəɹ/. Just because (in rule-based terms) flapping counterbleeds raising, that doesn't mean there's a phonemic distinction between /aɪ/ and /əɪ/. --Angr/_{tɔk tə mi} 20:06, 25 July 2005 (UTC)

That would depend on your theory of phonology assuming rule-based transformations. A debatable point, perhaps instead a question of morphophonology. A rider in a legislative bill has already become uncoupled from the verb ride (at least for many people); if a use of writer were to do something similar, it would be hard to argue for an underlying distinction in the consonant. Like the German fricatives, this is a marginal situation. My point was not that both should be included, as I doubt that would be useful for users of this article, but rather that the decision to exclude was debatable. kwami 20:47, 2005 July 25 (UTC)

There are lots of theoretical ways of achieving this kind of opacity; I gave the rule-based one as an example, although if I were writing a paper on this myself I would certainly use Optimality Theory and output-output faithfulness; someone else might use sympathy theory. That a legislative rider is semantically disassociated from ride doesn't matter, because rider is the one that has the vowel we'd expect before a flap. Only a meaning of writer that was semantically disconnected from write would be evidence for (but still not proof of) a phonemic split into /aɪ/ and /əɪ/, but even then since flapping is optional in American English ([ˈɹəɪtɚ] is not ungrammatical in AmE), there would still be evidence that such an independent meaning of writer had underlying /t/. And I do think this is an important phenomenon of English phonology that deserves to be reported on, but not here. This should be discussed at English phonology, not on a page that's supposed to be on how to use the IPA to represent English. --Angr/_{tɔk tə mi} 21:41, 25 July 2005 (UTC)

I wasn't trying to have it included here, just commenting. kwami

Audio clips?

Anyone interested in adding short audio clips as a learning aid? Gchriss 17:32, 3 August 2005 (UTC)

Secondary stress

I removed this from the suprasegmentals section:

English does not actually have a distinction between primary and secondary stress. The apparent difference is due to intonation: When making a statement, the last stressed syllable will be more strongly stressed than the other stressed syllables. However, as soon as you move a word out of final position, the extra stress is lost. It moves to whichever word is now final, so it doesn't really belong to the word itself, but to the statement. Consider the isolated word Arachnophobia, with stronger stress on the syllable pho than on the rach, versus Arachnophobia's playing at the Bijou, where the stress on rach and pho is equal. Because people usually say a word in isolation when transcribing it, they tend to mark primary and secondary stress, but this is not necessary for English.

This is patently false. There are plenty of words that have a stress pattern with secondary stress after primary stress, like in compounds, etc., and the latter part about stress belonging to the statement is just nonsense. Nohat 22:26, 20 August 2005 (UTC)

Patently false? Give us an example of contrastive stress levels in English. Sorry, Nohat, but I'd go with Ladefoged's opinion, even if I didn't have native speaker intuition that accords with this. (Haven't we had this argument before?) The distinction between primary and secondary stress in English (at least in most dialects) is intonational, as is all stress in French. It's just as wrong to say that English words have an inherent distinction between two types of stress as it is to say that French words are stressed on their final syllables. Here's what Ladefoged has to say on Degrees of Stress in his primer A Course in Phonetics (in this case the 3rd edition):

In some longer words, it might seem as if there is more than one degree of stress. For example, say the word "multiplication" and try to tap on the stressed syllables. You will find that you can tap on the first and the fourth syllables of "'multipli'cation." The fourth syllable seems to have a higher degree of stress. The same is true of other long words such as "'magmifi'cation" and "'psycholin'guistics." But this apparently higher degree of stress on the later syllable only occurs when the word is said in isolation or at the end of a phrase. Try saying a sentence such as "The 'psycholin'guistics 'course was 'fun." If you tap on each stressed syllable, you will find that there is no difference between the first and fourth syllables of "psycholinguistics." If you have a higher degree of stress on the fourth syllable in "psycholinguistics," this word will be given a special emphasis, as though you were contrasting some other psychology course with a psycholinguistics course. The same is true of the word "magnification" in a sentence such as "The de'gree of 'magnifi'cation de'pends on the 'power of the 'lens." The word "magnification" will not have a larger stress on the fourth syllable as long as you do not break the sentence into two parts and leave this word at the end of the first tone group.

Why does it seem as if there are two degrees of stress in a word when it occurs at the end of a phrase or when it is said alone—which is, of course, at the end of a phrase? The answer is that in these circumstances another factor is present. As we have seen, the last stressed syllable in a tone group usually carries the tonic accent. In longer words containing two stresses, the apparent difference in the levels of the first and the second stress is really due to the superimposition of an intonation pattern. When these words occur within a sentence in a position where there are no intonation effects, then there are no differences in the stress levels.

Table 5.3

Three-syllable words exemplifying the difference between an unreduced vowel in the final syllable (first column) and a reduced vowel in the final syllable (second column).

	'multiply	'multiple
	'regulate	'regular
	'copulate	'copula
	'circulate	'circular
	'criticize	'critical
	'minimize	'minimal

A lower level of stress may also seem to occur in some English words. Compare the words in the two columns in Table 5.3. The words in both columns have the stress on the first syllable. The words in the first column might seem to have a second, weaker, stress on the last syllable as well, but this is not so. The words in the first column differ from those in the second by having a full vowel in the final syllable. This vowel is always longer than the reduced vowel—usually [ə]—in the final syllable of the words in the second column. The result is that there is a difference in the rhythm of the two sets of words. This is due to a difference in the vowels that are present; it is not a difference in stress. There is not a strong increase in respiratory activity on the last syllable of the words in the first column. Both sets of words have increases in respiratory activity only on the first syllable.

In summary, we can say that English syllables are either stressed or unstressed. If they are stressed, they may or they may not be the tonic sylables, which carry the major pitch change in the tone group. If they are unstressed, they may or may not have a reduced vowel.

[...]

Some other books do not make the distinctions described here, maintaining instead that there are several levels of stress in English. The greatest degree of stress is called stress level one, the next is level two, the next level three, a lower level still is level four, and so on.

Ladefoged then goes on to explain that stress + tonic accent = level 1 in this system, that stress without tonic accent is level 2, that an unstressed unreduced vowel is level 3, and a reduced vowel is level 4, and then describes systems where there is no level 2 within words, so that the non-tonic stress levels become 3, 4, 5.

He continues,

I personally do not consider it useful to think of stress in terms of a multilevel system. Descriptions of this sort are not in accord with the phonological facts. But as it is so commonly said that there are many levels of stress in English, I thought I should explain how these terms are used. In this book, however, we will continue to regard stress as something that either does or does not occur on a syllable in English, and we will view vowel reduction and intonation as separate processes.

We can sometimes predict by rules whether a vowel will be reduced to [ə] or not. For example, we can formalize a rule stating that [ɔɪ] never reduces. But other cases seem to be a matter of how recently the word came into common use. Factors of this sort seem to be the reason why there should be reduced vowels at the end of "postman," "bacon," and "gentleman," but not at the end of "mailman," "moron," and "superman."

That is, secondary stress is traditional in descriptions of English, but is "not in accord with the phonological facts". (Compound words, which L does not go into here, add some complications, but do not add another phonemic stress level.) kwami 00:53, 2005 August 21 (UTC)

Just because Ladefoged prefers to consider secondary stress as part of "intonation" doesn't mean that there is no distinction between primary and secondary stress in English. This theory, by Ladefoged's own admission ("it is so commonly said that there are many levels of stress in English"), is contrary to the received wisdom of how English stress works. While there may be theories, even ones proposed by renowned phoneticians, that it is not necessary to posit an underlying multi-level system of stress at the lexical level, to assert such theories as fact, as you have done, is blatantly ignoring NPOV.

Furthermore, this is not a page that only represents a phonological view of English—the title itself says phonetic, and the contents already discuss many subphonemic aspects of English, of which multiple stress levels are part. There is a reason that all English dictionary makers mark multiple levels of stress in English—to do otherwise would make it impossible to know how to correctly pronounce words that have multiple stress! Please let this page simply describe the various IPA symbols that are used for transcribing English without bothering our readers with your marginally-relevant minority views on English stress, which would be better suited to English phonology or some such page.

Finally, the way you have presented this theory is inadequate because it fails to account for compounds like phrasal adjectives. You claim that compounds do not add another phonemic stress level but provide no reason why. The difference between new-class members and new class members is exactly the distinction between primary and secondary stress, and the fact that there are examples of contrastive stress levels in English makes it moot whether it is lexical or part of this vague and undefined notion of "intonation". It is true that the distinction between primary and secondary stress is levelled in many cases in English, but when a word does receive phrasal stress, a speaker has to know which of a word's stressed syllables will receive primary stress. The fact of the matter is multiple levels of stress are necessary to adequately describe spoken English. Nohat 04:57, 21 August 2005 (UTC)

Hi Nohat,

You wrote, Nevertheless, this theory does not account for cases where the position of primary stress is not predictable, as in compounds. Can you give an example of a compound where the position of primary stress is not predictable? kwami 04:27, 2005 August 21 (UTC)

See my example above. Nohat 04:57, 21 August 2005 (UTC)

Point taken about this belonging in the phonology article. Also, your NPOV edit is an improvement on what I wrote, except for that last comment.

But your example illustrates phrasal stress, not lexical stress. So it's not a counter-example to the claim that the primary-secondary distinction is intonational. kwami 05:48, 2005 August 21 (UTC)

I'm not clear exactly what "the claim that the primary–secondary distinction is intonational" means. My understanding is that secondary stress needs to be explicitly specified in the lexicon because otherwise it is impossible to know which stressed syllable in a word receives phrasal stress when the word receives phrasal stress. Your comments here and in the article seem to claim that it is always the final stressed syllable which receives primary stress, but this is plainly not true: words like stratosphere, westernmost, thunderstorm, synthesizer, etc. all have multiple stressed syllables, but don't have primary stress on the final stressed syllable. The same concern applies to my example: if there is only one level of stress, how do you account for the difference between new-class members and new class members in the underlying representation? It is the distinction between primary and secondary stress at the lexical level that ultimately accounts for the difference in phrasal stress. How do you account for which stressed syllable receives phrasal stress if you only posit one level of stress? Nohat 06:11, 21 August 2005 (UTC)

The lexical examples you give, stratosphere, westernmost, thunderstorm, synthesizer, all have lexical stress on the first syllable. What you're hearing as secondary stress are unstressed but full vowels. This is the point L made with Table 5.3, and is level 3 on his multi-level scale. Synthesizer, for example, would be [ˈsɪn.θə.saɪ.zɹ]. You may or may not care for this transcription, but it works: it provides all the information a native speaker needs to pronounce the word correctly: The stress mark shows where the tonic stress will dock, since it's the final stress. The unreduced vowel in [saɪ] show that this syllable is articulated clearly. The schwa and syllabic ar show that these syllables are reduced.

You can assume four stress levels in English (1ary, 2ary, 3ary, none) and by doing so you can cease worrying about the reduced vowels and tonic accent. Or you can assume three levels and disregard either the reduced vowels or the tonic accent. But if you choose to recognize tonic accent (which you have to do anyway, for phrasal and clausal intonation) and also recognize reduced vowels, then you no longer need anything more than [+stress] and [-stress].

(You were of course right in presenting this as one approach rather than the only one.)

As for the new-class members example, I don't know what stress you give this. Presumably you give it some sort of intonantion to disambiguate it from new class members. I can think of a couple ways to do this. That is, the stress isn't lexicalized, but is part of an intonation strategy, along the lines of the emphatic intonation Ladefoged mentioned: Tonic accent is normally on the last stressed syllable of a tone group, but it can be shifted for emphasis and disambiguation: "I want you to go", "I want you to go", "I want you to go", "I want you to go". However, it is always on the last stressed syllable of the word it docks to. This is the kind of thing that is going on with new-class members vs. new class members. We could count these as different types of stress, or even phonemes; this would even be a reasonable thing to do if people could agree on what should be included. But since this is very difficult, it is more usual to distinguish lexical phonetics from phrasal or clausal phonetics, and only speak of suprasegmental phonemes like stress and tone at the lexical level. And this article deals with the lexical level.

I've had this argument with a lot of people, because people are traditionally taught that English has a phonemic distinction between primary and secondary stress. No one has ever been able to give me a lexical item where simply marking [+stress] would cause any ambiguity, as long as reduced vowels are indicated. I even had a professor of phonology challenge me on this, and he was unable to come up with a single counter example. He conceded that, if you take vowel reduction to be part of the information stored in our mental lexicons (as Joan Bybee would argue), then there is no need for secondary stress. Of course, it's another question whether you want to do this; plenty of phonologists believe that vowel reduction is a product of stress, rather than vice versa.

Also, Ladefoged's spent a great deal of time investigating the acoustics and articulation of stress. If he says that syllables such as the [saɪ] in synthesizer do not have the hallmarks of stress that you get in syllables such as the [ˈsɪn] of synthesizer, then I have confidence that he knows what he's talking about, or at least more than most do. (Not, of course, that that necessarily makes him right.) kwami 07:12, 2005 August 21 (UTC)

Yeah, I've heard some of these arguments before but I have trouble buying them. To me, unstressed but unreduced vowels are vowels that don't count for timing. These are the vowels like the /i/ in happy, the /o/ in limo, the /u/ in influenza. I have trouble conflating those vowels with vowels in syllables that do count for timing, like the /saɪ/ in synthesizer. It's hard to account for a theory of timing unless you can distinguish which syllables count for it. Also, I don't think that either theory has any benefit from the standpoint of reducing information burden: the information used for intonation has to be stored somewhere. I personally find it difficult to believe that the fact that /seɪ/ is stressed in sensational /sɛnˈseɪʃənəl/ has absolutely no bearing at all whatsoever on the realization of sensationalistic /sɛnˌseɪʃənəˈlɪstɪk/. I can also tell you from personal experience that if you design a concatenative speech synthesizer to not distinguish between zero stress and secondary stress syllables when performing unit selection, the output is significantly less natural-sounding than that from a synthesizer which does, which is pretty strong evidence to me that there must be something to secondary stress. I also have trouble with the argument that vowel reduction necessarily wholly represented in the mental lexicon because it basically completely ignores the fact that literate people, when asked to "fully and carefully" pronounce a word, can unreduce pretty much any reduced vowel based on the spelling. See, for example, Robbie Williams' pronunciation of "melody" as /ˈmɛˌloʊˌdi/ in his performance of "De-Lovely" on the soundtrack to that movie. You might disregard that as spurious, but speech errors like that don't materialize from nowhere. It is clear that the underlying phonetic representations of words and the memorized spellings of words are deeply intertwined. Nohat 09:50, 21 August 2005 (UTC)

The /seɪ/ is stressed in both /sɛnˈseɪʃənəl/ and /sɛnˈseɪʃənəˈlɪstɪk/, so timing isn't a problem (with this example, anyway). This is different from the tertiary stress at the end of happy. Of course a speech synthesizer would need to distinguish between zero and secondary stress one way or the other; it would presumably (I don't know personally, since I've never worked with speech synthesis) need stress, tonic accent, and full vs. reduced vowels. Or you could go with primary through quaternary stress if you prefer. The point is that tonic accent and vowel reduction plus binary stress should produce all contrasts in English. (Although, of course, there are different types of tonic accent for different types of intonation.)

Joan Bybee has interesting thoughts on the mental representation of vowel reduction: every [ev.ri], memory [mem.ri] or [mem.r.i], emery [em.r.i], mammary [ma.mə.ri], for example, where the presence or absence of schwa and syllabic ar appear to be a function of frequency. (L mades the same point above.)

We need to be careful of claims based on the speech of literate people: Literacy definitely affects our perception of words. An illiterate person would have no idea that there's supposed to be an o in melody, unless of course they'd heard someone like Robbie Williams. So we have two parameters: mental representations of language based on speech, and representations based on writing, which are to a large extent conflated, of course. Even with its rather simple phonology, I've heard native Japanese speakers wonder whether two homonyms are pronounced differently, even when they have the same pitch accent, because they're written differently. kwami 20:15, 2005 August 21 (UTC)

OK. What you seem to be saying is that if we create a mapping from secondary stress to your system, syllables that have secondary stress before the primary stress in a word get [+stress] but syllables that have secondary stress after the primary stress get [-stress]. Is that right? Nohat 21:51, 21 August 2005 (UTC)

Perhaps. I don't know if that will always work (in lexical compounds, for example), since I've never been trained in a multilevel stress analysis. The way L correlated the two systems, you ask: Is the vowel full? does it have lexical stress? does it take tonic accent? If the answer is yes to all, it has primary stress; if yes to the first two, it is secondary; if yes only to the first, tertiary; and if the answer is no to all, it is quarternary. In a word like happy, I think a multilevel stress assignment would be 1.3; in sensational, it would be 3.1.4.4; and in sensationalistic it would be 3.2.4.4.1.3. Of course, a 'tone group' is defined intonantionally, so there will be all kinds of complications there when you get into real discourse. But if you frame your token so that the tonic accent falls after it, all the stressed syllables should level out, and are easier to distinguish from unstressed full vowels. I think that if you assign every syllable that counts for timing a 2, and then give the last one in a tone group a 1, the two approaches will predict the same results. kwami 00:12, 2005 August 22 (UTC)

I just meant the system of primary and secondary stress used by pretty much every English dictionary. For the examples I gave with secondary stress before primary stress, like sensationalistic, whereas dictionaries would specify /sen.sa-tion-a'lis-tic/ you specify /sen'sa-tion-a'lis-tic/. But for the cases where secondary stress came after the primary stress, like synthesizer, whereas dictionaries would specify /'syn-the.si-zer/, you would just specify /'syn-the-si-zer/. I'm guessing then that makes /co.a-gu'la-tion/ -> /co'a-gu'la-tion/; /.te-tra'he-dron/ -> /'te-tra'he-dron/; but /'thun-der.show-er/ -> /'thun-der-show-er/; /'boar-ding.hou-ses/ -> /'boar-ding-hou-ses/. (' marks primary stress on following syllable; . marks secondary stress; - or nothing marks no stress). The generalization I made from this is that when a dictionary specifies secondary stress, that translates into your system differently, depending on whether what the dictionary calls secondary stress precedes or follows primary stress. The assertion is essentially that what the dictionary marks as "the same thing" (secondary stress) isn't actually the same thing, and in fact the realization of those syllables depends on whether they precede or follow the primary stress. I'm not sure what I think of this theory yet, but this observation is troubling. Am I understanding correctly? Nohat 09:18, 22 August 2005 (UTC)

Yes, I believe you're correct. I haven't seen any exceptions to this pattern so far.

However, not all dictionaries transcribe stress this way consistently. The OED, for example, does not mark tertiary stress in a final syllable; rather, it simply gives a full vowel. For example, /'thunder.stricken/, but /'thunderstruck/. Also /'boardingschool/, not /'boarding.school/. It also does not mark 2ary stress immediately after 1ary; the full vowel again suffices: /'boardsailing/, not /'board.sailing/. I have yet to see an example where such 2ary stress coming after 1ary stress is unpredictable. It seems instead that 3ary stress is marked overtly when there are at least two two-syllable feet, and one of them would otherwise be without stress marking. In other words, the dictionaries are marking metrical patterns, perhaps for ease of scansion, but this is redundant as long as vowel quality is marked as well. (After all, dictionary pronunciation guides are meant to be as user friendly as possible; phonological consistency is a secondary concern. This could also be a tradition derived from reading poetry.) kwami 19:44, 2005 August 22 (UTC)

P.S. L's tried to define stress articulatorily, as he puts it, as "a strong increase in respiratory activity". This is not an easy thing to work on, but he apparently believes that dictionary 2ary stresses before the 1ary have this characteristic, but that dictionary 2ary stresses after the 1ary do not. That is, there is a physical difference at work here, not just a theoretical one. When taken out of tonic stress position, the 1ary stress levels with all the 2ary stresses before it, but remains distinct from the 2ary stresses after it -- what many phoneticians who recognize this effect call 3ary stress. That is, while /co'a-gu'la-tion/ has two equal stresses in non-tonic position, /'boar-ding.hou-ses/ does not, and therefore not all dictionary secondary stresses are equal. Also, by demoting these to full syllables rather than stressed syllables, it is predictable where the tonic stress will dock when the word does receive it: the final stressed syllable.

I guess you could also have two kinds of stress, primary and secondary, but /co'a-gu'la-tion/ would have two primary stresses, while /'boar-ding.hou-ses/ would have a primary and a secondary. Tonic stress would dock to the final primary stress, and unstressed vowels would be reduced. This might depend on whether you want stress or full vowels/heavy syllables to drive your metrical phonology. kwami 23:19, 2005 August 22 (UTC)

Unicode code?

The HTML code for numeric entities for the Unicode code

Is this really necessary? —Frungi 22:26, 26 August 2005 (UTC)

I've improved the language, but if you mean is it necessary to show the codes at all, I would agree that it doesn't belong in this article. Anyone who would want the HTML entity for a character should be able to use the actual Unicode character itself just as easily. If someone needs a Unicode-entity reference, this article isn't the place for it. I'll remove it and see if anyone objects. —Michael Z. 2005-08-27 16:50 Z

Canadian English

Does anyone care to add Canadian English to this article? There are many differences from U.S., but I think they're too subtle for me to try to document here. —Michael Z. 2005-08-27 16:59 Z

I could possibly do it, if nobody else wants to.--Sonjaaa 19:03, August 27, 2005 (UTC)

I'll pitch in where I can, or do a first draft of additions to the table, anticipating your corrections and additions. —Michael Z. 2005-08-28 01:40 Z

I'm agin it. This page is not the place to discuss the differences between dialects of English. There are plenty of other pages where that can be discussed (e.g. North American English, Canadian English). The focus of this page should be on how the IPA is used in pronunciation guides of Wikipedia articles and should be kept simple. --Angr/_{tɔk tə mi} 18:41, 1 September 2005 (UTC)

New content

Well, no one has objected, so I'm replacing the content of this page with the content of User:Angr/IPA for English (proposal). --Angr/_{tɔk tə mi} 08:27, 5 September 2005 (UTC)

graphic representations of mouth and tongue posistions

If this is the appropriate place for simple line art images depicting tongue, lip, teeth and mouth positions for all the phoneme parts, will someone please create some and add them here? (or find a licence compatible source and import them)

Such as http://www.sk.com.br/mouth-th.gif for "th" sound in the, thirty, them, Thursday, and the like.

Thanks - Scott Edwards

I don't think this article is the appropriate place for such images. Place of articulation has a general picture of the vocal tract. Perhaps the articles on individual sounds, like Voiceless postalveolar fricative, would be improved by line drawings of how they're produced. --Angr/_{tɔk tə mi} 06:12, 7 October 2005 (UTC)

Encoding and fonts

I can't see most of the IPA symbols on this wiki page. I'm using IE 6 and have encoding set to Unicode (UTF-8). 2005-10-21 18:59:00 (PST)

That's annoying, because they all use the {{IPA}} template, which was supposed to solve the problem of people using IE. All I can suggest is use a different browser! --Angr/_{tɔk tə mi} 02:18, 22 October 2005 (UTC)