My own analysis of my English's vowel system

[Travis B.]

One should remember that quite a few English varieties do normally have monophthongs for historical /eɪ̯/ and/or /oʊ̯/, which consequently are often marked as simply /e/ and /o/ respectively. Also, at least in General American (aside from more progressive varieties thereof that have lost historical vowel length) and Received Pronunciation, /ɑː/ is indeed long and not short.

A finlay points out, this is how I talk. fate and boat are roughly [fet], [bot]. This doesn't change the fact that for General American, these vowels are transcribed as /eɪ/ and /oʊ/.

In North American English contexts, I still see these not infrequently referred to as /e/ and /o/ rather than /eɪ̯/ and /oʊ̯/, even though I would say that the latter pair are more correct for speaking about NAE as a whole.

I prefer to use /e/ and /o/, mainly for notational convenience; for me they vary between pure vowels and diphthongs, and not always predictably. I would find it strange to refer to weather [veinz], but the blood vessels can be either [venz] or [veinz]. There are historical reasons for this, but I have no idea why they should have affected my own idiolect.

For me, these are normally monophthongs unless followed by another vowel or, when I am approximating General American, word-final, where then they may be diphthongs, but only are so very inconsistently. I tend to prefer to analyze my dialect as having these as monophthongs, with cases in which they have glides synchronically being cases of optional epenthesis, especially since how frequently these glides appear depends heavily on not only register but also which vowel follows them.

However, for speaking about North American English as a whole, I still prefer to speak of /eɪ̯/ and /oʊ̯/ rather than /e/ and /o/, no matter how I analyze my own dialect, due to this being more diachronically correct and broadly applicable. Even in my own dialect, the monophthongs I have therefor are almost certainly descended from a historical /eɪ̯/ and /oʊ̯/, with the cases where I do have off-glides diachronically not being cases where glides were inserted but rather cases where glides were not lost*.

* This is especially since I have certain words such as cooperate [kʰəːˈwapʁ̩ˤːʁˤeʔ] (1) where the offglide was preserved as a distinct segment /j/ or /w/, and not subject to further reduction of the /eɪ̯/ and /oʊ̯/ from which they originated.

(1) X-SAMPA: [k_h@:"wapR=_?\:R_?\e?]

[Rory]

There are Anglic dialects that do have a true diphthong phoneme /i̯u/ or /iu̯/ from Middle English /iu̯/ and /ɛu̯/ and Old Norman, Old French, and Middle French /y/, before which they actually do use an (to the best of my knowledge). Mind you that in these dialects, there are still words with /juː/, such as youth, and these words still would not take an.

What are these dialects? What is the phonetic difference between a "true diphthong" and a sequence of glide+vowel? How can we prove that these dialects have a phonological distinction between /i̯u/ and /juː/? I want to believe you, but you have to bring evidence to support your claims.

The dialects I've seen something like this described for are in Wales, e.g. Rhondda (PDF; the relevant section is 3.14). Note though that he says that words with RP initial /ju:/ have an initial [j], so maybe phonologically they start /jIU/ and would be expected to take "a".

Thanks! That dialect looks very interesting, I'd love to get a listen to it.

One should remember that quite a few English varieties do normally have monophthongs for historical /eɪ̯/ and/or /oʊ̯/, which consequently are often marked as simply /e/ and /o/ respectively. Also, at least in General American (aside from more progressive varieties thereof that have lost historical vowel length) and Received Pronunciation, /ɑː/ is indeed long and not short.

A finlay points out, this is how I talk. fate and boat are roughly [fet], [bot]. This doesn't change the fact that for General American, these vowels are transcribed as /eɪ/ and /oʊ/.

In North American English contexts, I still see these not infrequently referred to as /e/ and /o/ rather than /eɪ̯/ and /oʊ̯/, even though I would say that the latter pair are more correct for speaking about NAE as a whole.

Yep, although you never see them as /oː/, which the OP suggested. I'm not trying to impose a "correct" transcription for GA, just saying what is most common, and I thought that the OP could benefit from that.

In RP, /ɑː/ represents a different lexical set than General American /ɑ/ (BATH vs COT). I'm aware that different people speak in different ways. The OP was contrasting the conventional GA analysis with his own. I was correcting aspects of the GA analysis that he missed or got wrong.

Okay, cases of RP /ɑː/ generally cannot be said to be uniformly belong to the same lexical set as GA /ɑː/, except in words like father, palm, and start. The GA vowel is still historically long rather than short, though.

Well, the conventional analysis of GA is that it has /ɑ/, not /ɑː/. (Palm is also different as many GA speakers pronounce the /l/, whereas that is not done in RP.) Are you sure the GA vowel is historically long? I'm not so sure, given that /ɑː/ did not exist in vowel systems on either side of the Atlantic when the US was first being settled (as far as I know).

There are Anglic dialects that do have a true diphthong phoneme /i̯u/ or /iu̯/ from Middle English /iu̯/ and /ɛu̯/ and Old Norman, Old French, and Middle French /y/, before which they actually do use an (to the best of my knowledge). Mind you that in these dialects, there are still words with /juː/, such as youth, and these words still would not take an.

What are these dialects? What is the phonetic difference between a "true diphthong" and a sequence of glide+vowel? How can we prove that these dialects have a phonological distinction between /i̯u/ and /juː/? I want to believe you, but you have to bring evidence to support your claims.

Now that I think of it, I cannot say that such a contrast exists offhand, that is that initial /i̯u/~/iu̯/ is not treated differently from medial /i̯u/~/iu̯/ and do not have anything like /j/ inserted before them. I could dig for information here, but really cannot do so at the very moment. However, there are cases of /juː/ which clearly are not cases of /i̯u/~/iu̯/ diachronically, youth being one of them.

Yeah, youth clearly is historically a cluster, whereas cute is historically a single vowel. However, in the absence of a contrast, I can't see any evidence for claiming that they are distinct categories - at least not any more than the vowels in meet and meat are distinct categories (in GA).

On a different topic entirely: Travis, do you have any recordings of yourself anywhere? I'd love to hear what you sound like, as you describe it with such precision.

[finlay]

We've tried (well, I did). He came up with one once, but then when it didn't back up any of the claims he makes in his transcriptions (which he still hasn't learnt to moderate for relevance and expected level of the recipient), he went on to claim that it wasn't representative of his dialect. Too formal or for too wide an audience. He still hasn't come up with the goods.

I still claim observers paradox on that ʁ until I hear it. No coronal articulation, sure, but I'm not so sure about ʁ.

[Nortaneous]

We've tried (well, I did). He came up with one once, but then when it didn't back up any of the claims he makes in his transcriptions (which he still hasn't learnt to moderate for relevance and expected level of the recipient), he went on to claim that it wasn't representative of his dialect. Too formal or for too wide an audience. He still hasn't come up with the goods.

Not that surprising; I've tried to make recordings of my idiolect, but if I read a prewritten text, it comes out in something much closer to GA than what I actually speak.

[Travis B.]

We've tried (well, I did). He came up with one once, but then when it didn't back up any of the claims he makes in his transcriptions (which he still hasn't learnt to moderate for relevance and expected level of the recipient), he went on to claim that it wasn't representative of his dialect. Too formal or for too wide an audience. He still hasn't come up with the goods.

I was speaking rather carefully, IIRC, because I had just gotten a new but still very cheap microphone, and I did not know how well it would pick up my voice, as I have had very bad luck with cheap computer microphones in the past. Also, I was not speaking like I would normally to people I know personally back in Wisconsin but rather like how I would speak to a stranger on the phone.

Aside from that, a lot of the time I honestly cannot be bothered to make recordings, much the less good ones, at a drop of the hat, because it is much more work than it sounds to do well, much the less in a way that is representative of my everyday speech with people I know, and a lot of the time I really am not up to it.

I still claim observers paradox on that ʁ until I hear it. No coronal articulation, sure, but I'm not so sure about ʁ.

The reason why you say that is that it is not uvular all the time; in syllable onsets it has a velar POA or, after a coronal, coarticulated velar and postalveolar POAs.

[finlay]

..

observers' paradox: the bane of linguistics. also reading aloud and talking shit are very different skills, it is true. but it's part of the reason that you don't as a general rule do linguistic analysis on yourself, even though it's so tempting and you have all the articulatory data right there, but you can't be sure that it's accurate because you're thinking about it too much.

[TaylorS]

My /e/ and /o/ are pure monophthongs, not glide to them at all.

My vowel inventory is:

iː ɪ ʊ uː
eː ɛ ʌ oː
æ a
aɪ ɑʊ ɔɪ
ɹ̩ l̩ m̩ n̩

I am Caught-Cot Merged.

/æ/ is realized as [ɛ] in unstressed syllables, [ɛə] elsewhere.

/aɪ ɑʊ/ undergo Canadian Raising.

/eː/ and /oː/ are mid-vowels, not high-mid

/ɪ ʊ/ reduce to [ɨ̞] when unstressed, all others except /æ/ reduce to [ə]

The frontness or backness of /a/ is determined by the adjacent consonants, otherwise it is by default central.

/uː/ fronts to [ʉː] when following coronal consonants.

[makvas]

Well then my dialect only has six vowels: /A { E I @ U/.
Everything else is a rising diphthong: /Aj {w Ej Ij @w Uw/ and if you count those syllabic stuff: /@r @l @m @n/.

[Rory]

Aside from that, a lot of the time I honestly cannot be bothered to make recordings, much the less good ones, at a drop of the hat, because it is much more work than it sounds to do well, much the less in a way that is representative of my everyday speech with people I know, and a lot of the time I really am not up to it.

Do you have a smartphone? Many of them are able to record telephone conversations, you could do that next time you call someone from Wisconsin - obviously edit out the other end of the conversation, and remove any personal information. Or if your computer has a microphone, just set it to record if you're likely to be staying in the same place but doing a lot of talking.
Although you're right, getting good quality recordings of spontaneous speech is hard. And even perfect recordings wouldn't be able to prove or disprove any theories about the coronality or uvularity of your /r/, for instance - to be sure of that, we'd have to use an ultrasound or some other sort of advanced articulatory imaging technology.

[Bedelato]

Ewe/yew/you sounds like it would be a good way to test if any dialects have a true contrast between /ju/ and /iu/. Does any such dialect exist?

Yeah, I contrast ewe/you.

In RP, /ɑː/ represents a different lexical set than General American /ɑ/ (BATH vs COT). I'm aware that different people speak in different ways. The OP was contrasting the conventional GA analysis with his own. I was correcting aspects of the GA analysis that he missed or got wrong.

Okay, cases of RP /ɑː/ generally cannot be said to be uniformly belong to the same lexical set as GA /ɑː/, except in words like father, palm, and start. The GA vowel is still historically long rather than short, though.

Well, the conventional analysis of GA is that it has /ɑ/, not /ɑː/. (Palm is also different as many GA speakers pronounce the /l/, whereas that is not done in RP.) Are you sure the GA vowel is historically long? I'm not so sure, given that /ɑː/ did not exist in vowel systems on either side of the Atlantic when the US was first being settled (as far as I know).

The /ɑ:/ thing in my original post was a typo. I don't know how that long marker slipped in there.

Rory's said anything that I would... I'd add, though, just explicitly that your analysis is closer to what I'd normally see about GA anyway, apart from that I've never seen the three-low-vowel thing either. (are you sure one of them's not rounded, for instance, even if it's not "fully" rounded?) Also, when talking about phonemes, "lousy fit" doesn't really matter: they're arbitrary symbols. It does depend to some extent on the purpose of the transcription, to be fair, but the point of, say, the phonemes given on wikipedia is to somehow accomodate all the dialects of English.

Also, careful with brackets:
"As a rule, /u/→/ʉ/, except before /l/, as in "coo" [kʰʉ:] vs. "cool" [kʰuɫ]."
/u/-›[ʉ], it should be. Weirdly, you then get it right a second later. As an aside, I don't think I know any English speakers with as the default realisation of /u/ (although there's a lot of variation with this, and I think it happens more in America than Britain), and it's a good example of how arbitrary the symbols can be.

Oh boy, yet MORE nitpicking about brackets. I bet someone brings that up on every other thread here.

I've thought the same thing to those Indo-Europeanists. While they're scrapping with each other over "glottalic theory" stuff, I sit here saying, "WHO CARES!? The phonemes are just arbitrary symbols. What does it matter?" Their only real complaint is that voiced aspirates without voiceless counterparts are "rare". And frankly, original voiced aspirates make perfect sense with sound changes in mind.

One of them might be more rounded than the other. But there's still a three-way contrast, and it's not the un-father-bothered type. I do in fact merge father/bother, and my distinction is in a completely different lexical set.

[Whimemsz]

observers' paradox: the bane of linguistics. also reading aloud and talking shit are very different skills, it is true. but it's part of the reason that you don't as a general rule do linguistic analysis on yourself, even though it's so tempting and you have all the articulatory data right there, but you can't be sure that it's accurate because you're thinking about it too much.

Although you're right, getting good quality recordings of spontaneous speech is hard. And even perfect recordings wouldn't be able to prove or disprove any theories about the coronality or uvularity of your /r/, for instance - to be sure of that, we'd have to use an ultrasound or some other sort of advanced articulatory imaging technology.

And this gets to the heart of why the really ultra-detailed transcriptions are--intellectually-speaking--untenable. Without spectrograms and palatograms and so on, there's no way for anyone, no matter how linguistically-competent, to infallibly record the phonetics of their speech in that much detail. And in any case since at THAT level of transcription no one pronounces the same word identically from sentence to sentence anyway, it makes no sense to use the level of transcription Travis uses unless you're transcribing a specific, individual recorded statement or discourse or whatever.

[Bedelato]

... just explicitly that your analysis is closer to what I'd normally see about GA anyway ...

I'm guessing you meant "say"?

[Whimemsz]

Why? "See" works there: what he normally sees written/described about GA.

[Bedelato]

Oh.

[Bedelato]

Do you have a smartphone? Many of them are able to record telephone conversations, you could do that next time you call someone from Wisconsin - obviously edit out the other end of the conversation, and remove any personal information. Or if your computer has a microphone, just set it to record if you're likely to be staying in the same place but doing a lot of talking.
Although you're right, getting good quality recordings of spontaneous speech is hard. And even perfect recordings wouldn't be able to prove or disprove any theories about the coronality or uvularity of your /r/, for instance - to be sure of that, we'd have to use an ultrasound or some other sort of advanced articulatory imaging technology.

Don't you have to ask permission first to record a phone call?
I heard/read somewhere that it's illegal if you don't. I might be wrong, though...
But that would introduce observer's paradox

[Alces]

Are you sure the GA vowel is historically long? I'm not so sure, given that /ɑː/ did not exist in vowel systems on either side of the Atlantic when the US was first being settled (as far as I know).

Actually, it did exist in both vowel systems; at least in the cases of 'father', where ME /a:/ resisted the GVS, 'palm', where it results from L-vocalisation in /al/ > /au/ > /a:/ (before labials), 'start' where short vowels were lengthened before coda /r/, and 'spa' where it was in an environment that didn't allow short vowels. All of these changes make the most sense if it was originally long.

[Whimemsz]

It could work if, say, you tell your family members, "hey from now on for the foreseeable future I'll be recording our phone conversations to obtain phonetic data", then you wait for like three months until they've basically forgotten about it and start recording them then.

[Travis B.]

The problem with recording other people with their knowledge is they often change their pronunciations to sound "more correct" when you do it. Actually, people do that even if they think you are listening to them for, well, linguistic data. Of course, the implication of this is that getting reliable answers to a particular phonological question is much harder than it seems it would be, and that much the time the best way to get good information on how someone pronounces something is to simply wait for them to say it - but then you do not have a recording but only how you remembered them say it in normal everyday conversation without any recording or even attempting to "steer" them to say a particular word, which is typically no good for those who insist on spectrograms and whatnot.

[finlay]

Do you have a smartphone? Many of them are able to record telephone conversations, you could do that next time you call someone from Wisconsin - obviously edit out the other end of the conversation, and remove any personal information. Or if your computer has a microphone, just set it to record if you're likely to be staying in the same place but doing a lot of talking.
Although you're right, getting good quality recordings of spontaneous speech is hard. And even perfect recordings wouldn't be able to prove or disprove any theories about the coronality or uvularity of your /r/, for instance - to be sure of that, we'd have to use an ultrasound or some other sort of advanced articulatory imaging technology.

Don't you have to ask permission first to record a phone call?
I heard/read somewhere that it's illegal if you don't. I might be wrong, though...
But that would introduce observer's paradox

You can't do linguistic analysis without potential observers' paradox. Not legally anyway. The possibly exception is Labov's rapid and anonymous test where you ask somebody in the street (or in his case, a department store in New York) and record what they say. Basically half the research in sociolinguistics is finding out ways of reducing the effect.

Whimemsz's suggestion is, IIRC, a bit questionable. With your family members, you can get away with more, however.

[Bedelato]

Are you sure the GA vowel is historically long? I'm not so sure, given that /ɑː/ did not exist in vowel systems on either side of the Atlantic when the US was first being settled (as far as I know).

Actually, it did exist in both vowel systems; at least in the cases of 'father', where ME /a:/ resisted the GVS, 'palm', where it results from L-vocalisation in /al/ > /au/ > /a:/ (before labials), 'start' where short vowels were lengthened before coda /r/, and 'spa' where it was in an environment that didn't allow short vowels. All of these changes make the most sense if it was originally long.

I pronounce the /l/ in "palm", "walk", "talk".
It's not that audible, but the coronal articulation is still there.
I also ha

I have "start" [stʌɻt̚] with raising of /a/ before /ɻ/+voiceless consonant in the same syllable in the same morpheme.
Raising also occurs sporadically before /rd/. I have a minimal pair for this: "barred" [baɻd] vs. "bard" [bʌɻd]. "Lard", "guard" and "card" are other affected words, but "hard" isn't.
This might be related to /ai/ raising (see below).

A few more allophonic rules in my idiolect (in order where needed):

/ʌ/ > [ə] before alveolar and post-alveolar consonants, or when unstressed. (So it is wedge after all! Who knew?)

I have a "roses"~"Rosa's" contrast, but the environment is really restricted.
The distinction is pretty much restricted to initial and final syllables, collapsing almost everywhere else. It also collapses before nasals in the syllable coda, so "bacon" and "bakin'" are homophones, but "a mission" and "emission" are distinct, because the /m/ is part of the next syllable.

/l/ and /ɻ/ are pharyngealized in codas after back vowels.
/l/ is pharyngealized when syllabic.
/ɻ/ is labialized initially.
/ɻ/ is "bunched" in codas. (Question: What is the IPA for "bunched r" anyway?).

The vowel known in traditional English phonics as "long <o>" and transcribed by me as /əw/ is regularly realized as [œw].

I have raising of /ai/ to /əi/ before voiceless consonants. It's not Canadian raising, because /au/ isn't involved. There are also sporadic anomalies of raising before /d/: "hide", "spider", and the name of a certain terrorist organization. Question: Does the sporadic component make the split phonemic?

/t d/ assimilate their POA to a following /θ ð/.

"Partial th-stopping" (my moniker for it, because it's... conditioned fortition of <th> ):
-/θ/ becomes [t̪]~[tθ] (dental stop/affricate) after sibilants
-/ð/ becomes [d̪]~[dð] after any coronal.
There is free variation between stop and affricate for both /θ, ð/ here.

Most instances of "th-stopping" that I've heard about are unconditional. Mine isn't

My treatment of /æ/:
-Becomes [ɛ] when unstressed
-Becomes [e] before nasals
-[e] from above rule > [eə] before /m, n/.
-Nasalization of all vowels (including diphthongs) before nasals.

[Nortaneous]

/ɻ/ is "bunched" in codas. (Question: What is the IPA for "bunched r" anyway?).

you do not want to open that can of worms

just write it as ɹ

if you mean what is it phonetically, a quick google gives a description as a "voiced labial prevelar approximant with tongue-tip retraction"

[Whimemsz]

/ɻ/ is "bunched" in codas. (Question: What is the IPA for "bunched r" anyway?).

I don't think there's any universally-acknowledged way of indicating it, so you kind of have to go with some ad-hoc combinations. I personally usually go with [ɹ] with a "retracted" diacritic: [ɹ̠]. Except that if you're going to be that precise you might as well include shit like velarization/pharyngealization and rounding/labi(odent)alization...

Which is basically like what I was lecturing Travis about in the other thread so... :| .... awkward

[Bedelato]

There is actually velarization the way I say it (but it's pharyngealized instead before back vowels. I think I said that already, but I'll say it again for the heck of it).

And... labial? How the ____ is it labial?

Anyway, consonantal /ɻ/ is in fact retroflex for me. I'm 90% sure of that.

Is there a way I could replace "vowel system" with "phonology" in the topic title?

[makvas]

Just looking at *any* analysis of English, I have a feeling that you could pretend that English has postpositions in some dialects. Like in Southern US English,

one of the guys knows it <-- "correct"
one of the guys know it <-- more common

This suggests that "the guys" is the subject, "one of" would be the adpositional phrase, instead of "of the guys". This breaks down with any prepositional phrase that's not adjectival in nature, but it's still interesting.

[eodrakken]

Ewe/yew/you sounds like it would be a good way to test if any dialects have a true contrast between /ju/ and /iu/. Does any such dialect exist?

That is a very good question that I would be interesting in seeing the answer to myself.

What about the expression of disgust "ew"? I've certainly heard [iu] for that, and no one would confuse it with [ju]. Or does that not count?