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.

As far as I can see there isn't any strong reason to postulate a close-mid–open-mid contrast in GA. There are only three heights (close, mid, open) and three degrees of backness (front, central, back) and two degrees of tenseness (lax and close) for each phonological backness.

Both /oʊ/ and /ɔ/ are clearly tense (they can occur in a word-final position, both stressed and unstressed), and if /oʊ/ is truly a phonological monophthong then the two differ in height. There's nothing wrong with classifying /ɔ/ as open, in fact it's more logical to do so since that's the historic pronunciation and GA is known to be more archaic than RP. Classifying /ɑ/ as central is also correct as far as historical phonology is concerned (well, it would be even more correct to classify it as front as it clearly used to be a long counterpart of /æ/. But it's never phonetically front and unlike /æ/ it isn't raised before nasals) and it shows even better that it forms a natural class of tense monophthongs with /i, u, eɪ, oʊ/ (the only low vowel in German, Spanish and Italian is also central). When it comes to the cot–caught merger, it seems to me that it can be easily explained as /ɑ/ becoming [+back] (or perhaps [-central] [+back] if there indeed are more phonological central vowels than /ɑ/ in GA) as rounding is non-phonemic in GA.

I think that /a/ and /ɒ/ (or /aː/ and especially /ɒː/) are much better representations of COT and CAUGHT in GA, but we obviously need to follow the sources. Also, with the merger being common nowadays, ⟨ɑ⟩ or ⟨ɑː⟩ is one of the most appropriate symbols for the merged vowel, but I don't think that the neutralization should be understood as a disappearance of a contrastive THOUGHT vowel. The way I see it, /ɑ/ becomes [-central] [+back] and that way it merges with the open back /ɔ/. Roundedness is probably non-phonemic in GA and it's possible that /ʌ/ is actually [+central] [+low] in GA, as in Australian English. In that analysis, the phonetically unrounded /ɪ, i, ɛ, eɪ, ə, ɜ, æ, ʌ, ɑ/ can be differentiated from the phonetically rounded /ʊ, u, oʊ, ɔ/ by backness alone.

(Redacted)

It certainly plausible to analyze /ʌ/ as central because of the close relationship between it and /ə/. Word-finally, /ə/ is often close to the main allophone of /ʌ/ (which is [ɐ]) and the two don't always contrast in GA (they certainly don't at the end of a word). It's also logical to analyze /ʌ/ as the lax variant of /ɑ/ because the former is more mid in terms of height, as it is the case with the vowel pairs of close height.

(Redacted)

Currently, we say that /ɜ/ is not a true phoneme in General American but merely a different notation of /ə/ preserved for when this phoneme precedes /r/ and is stressed—a convention adopted in literature to facilitate comparisons with other accents. This is fair enough. But what about the length marks? Traditional RP also doesn't seem to feature phonemic vowel length (at least when you ignore the schwa) and all tense vowels are transcribed with different symbols anyway. Why don't we introduce the length marks to this article per LPD and CEPD? We're already inconsistent in transcribing the phonological mid tense /eɪ/ and /oʊ/ as such rather than using ⟨e⟩ and ⟨o⟩ for them. We're also using ⟨ɔ⟩ for a vowel that is more appropriately written with ⟨ɒ⟩ as it differs from the unrounded /ɑ/ mainly by backness and roundedness, rather than height and rounding.

Other pros of introducing the length marks for /i, u, ɜ, ɑ, ɔ/ include the fact that the notation will finally match Help:IPA/English and that ⟨i⟩ will be freed up to be used for the HAPPY vowel (which isn't the same as FLEECE - that's just an oversimplification). It'll also allow us to change ⟨ɔː⟩ to ⟨ɒː⟩ should we ever want to do so (some works on British and especially Welsh English use ⟨ɒː⟩ for the THOUGHT vowel, but ⟨ɒ⟩ is only ever used for the LOT vowel).

(Redacted) Kbb2 (ex. Mr KEBAB) ( talk) 15:14, 6 March 2019 (UTC)

I think that classifying /æ, ɑ, ɔ/ (actually /æ, a, ɒ/ or perhaps /æ, ɐː, ɒː/ if we want to be pedantic) as lax front, tense central and tense back illustrates perfectly the reason for which the last two tend to merge, and that reason is that many speakers of NAE strive to have a simpler and more natural vowel system and /ɔ/ is simply too strange to be a part of it. I'm pretty sure that in the minds of native speakers without the merger the vowel system of GA is something like this:

/ə/ stands for COMMA, STRUT and NURSE. Then after the merger it's more like:

There are obvious similarities between this and the /ɛː–eː/ merger in German. Kbb2 (ex. Mr KEBAB) ( talk) 15:56, 6 March 2019 (UTC)

I like your basic chart (with the one exception that I'd argue your /a/ phoneme to be /ɑ/ instead). And I've been a long-time supporter that the phonetic realization of unmerged GenAM /ɔː/ is [ɒː] (though I get that you're trying to make a phonemic rather than phonetic system). Anyway, amidst all your writing, I feel like I've missed any actual point you're trying to make. Are you suggesting we change something? Wolfdog ( talk) 13:57, 10 March 2019 (UTC)

@ Wolfdog: I wrote it with ⟨a⟩ because that's the most common transcription of the open central unrounded vowel.

I'd certainly support using ⟨ɒː⟩, but who writes THOUGHT like that besides the authors (I think that's actually Wells himself) of one of the Longman dictionaries? I guess nobody does. I wouldn't support writing it with ⟨ɒ⟩ because no source I'm aware of does that and because it clashes with the way LOT is written in transcriptions of British, New Zealand and South African English.

I'm not arguing for any of my tables to be included in the article, they're there because I'm trying to make a point (which I should've clarified, that's true). The main point is the introduction of the length mark ⟨ː⟩ for tense close and open vowels /i, u, ɜ, ɑ, ɔ/ to make comparison with other accents even easier and to match Help:IPA/English more closely. GA isn't Scottish English and writing /iː, uː, ɜː, ɑː, ɔː/ isn't wrong in any way, just somewhat redundant but it's also redundant in the case of traditional RP in which all of the tense vowels are transcribed with symbols different than those used for the lax vowels, even those that shouldn't (meaning the /ə–ɜː/ pair - the tense vowel should be written with ⟨əː⟩. If unrounding and lengthening of /ɒ/ triggered the loss of distinctive length in GA then perhaps it should be written with /ɑ/ as well - indeed, that's what Geoff Lindsey has been arguing for). GA is also unlike Standard Dutch, in which three of the seven native tense vowels are short (compare /i, y, u/ with /eː, øː, oː, aː/), though this applies only to stressed syllables. And even in SD you could probably argue that /i, y, u/ are underlyingly /iː, yː, uː/, but I don't want to go there now (nor should I).

We already write the NURSE vowel with ⟨ɜ⟩ and that is redundant because [ʌ] can't occur in that position, so it only makes sense to analyze NURSE as /ʌr/ (and we don't). We also write the tense mid vowels with ⟨eɪ, oʊ⟩ and that's inconsistent with both the other tense vowels of GA and the way RP vowels are transcribed (and also the way we transcribe English vowels on Help:IPA/English).

I think that a truly underlying vowel system of GA is actually this:

	Front		Central		Back
	short	long	short	long	short	long
Close	i	iː			u	uː
Mid	e	eː	ɐ			oː
Open	æ			aː		ɒː
Diphthongs	ai   ɔi   au

The centralized [ɪ, ʊ, ɛ] allophones of /i, u, e/ are triggered in phonologically closed syllables (making syllabification such as /ˈlɛ.tər/ or /ˈle.tɐr/ for letter 'wrong'). This might seem like a somewhat nonsensical statement but syllable- and especially word-final instances of /uː/ and /eː/ can be analyzed as /u/ and /e/ because [ʊ] and [ɛ] don't occur in this position. Following this logic, /æ/ should be analyzed as /a/, the short counterpart of /aː/. When it comes to /i/ and /iː/, see the first message.

(Redacted)

In stressed syllables, what we now write with /i, u, eɪ/ are always longer (often diphthongal) than /ɪ, ʊ, ɛ/ (/ɑ, ɔ/ typically have the length of the former two). I think that it's not very plausible to postulate a contrast between /ɛ, e, ɪ, i/ if /e/ and /i/ are always longer than /ɛ, ɪ/ (much as in Standard German and Swedish). The actual contrast is between /e, eː, i, iː/, even though native speakers of NAE are very likely to perceive close [e, i] used for /ɛ, ɪ/ as /eɪ, iː/, especially if one fails to produce American vowels with correct length. But phonology is not phonetics. It's abstract by definition and I think that there should be a difference between the phonological treatment of NAE and (broad) phonetic transcriptions for more practical purposes, much like what you can see in Duden's Das Aussprachewörterbuch (but even there there's room for improvement, though analyzing their transcriptions here would certainly get a chunk of my post removed per WP:NOTAFORUM).

I wonder if there's any English dialect that features a phonemic contrast between /ɛ/, /e/, /ɪ/ and /i/ with all of them having the same phonetic length in the same environment. Scottish English doesn't seem to be one, because in it /ɪ/ often has a schwa-like quality and you could argue that it forms a checked-free pair with /ʉ/ because there's no contrastive FOOT vowel in ScE. So I don't even think that it features four phonemic vowel heights; rather, it features close /i/, /ɨ/ (the KIT vowel), /ʉ/ and /o/ (the GOAT vowel), mid /ɛ/, /e/, /ɐ/ and /ɔ/ and an open /a/. It's not a perfect analysis (there are gaps), but it looks pretty solid to me.

I'm still curious how modern sources treat /ʌ/. Is it still a phonetically advanced back vowel in GA? Or has it become a simple [ɐ] for the majority of speakers?

tl;dr: Let's add length marks for /i, u, ɜ, ɑ, ɔ/ and then perhaps change ⟨ɔː⟩ to ⟨ɒː⟩ if we find enough sources for that. Kbb2 (ex. Mr KEBAB) ( talk) 16:20, 10 March 2019 (UTC)

@ Wolfdog and Nardog: Ok, another tl;dr (a longer one): the reason I want to add length marks for /i, u, ɜ, ɑ, ɔ/ is that ⟨iː, uː, ɜː, ɑː, ɔː⟩ match what we can find on Help:IPA/English and in many WP articles on English phonetics and phonology. They're not "wrong" (GA isn't Scottish or Ulster English) and accurately represent the phonetic realization of these vowels in that they're longer than /ɪ, ʊ, ɛ, æ/, though the lax open front one might be as long as /ɑː/ when it's tensed to [ɛə] (and perhaps also when it occurs before lenis consonants, as in RP). When it comes to /ɜː/ (actually /ɜːr/), it is simply /ʌr/ (or, when unstressed, /ər/) transcribed in a different manner. But ⟨ɜː⟩ is just as phonemically inaccurate as ⟨ɜ⟩ (both falsely indicate a phonemic distinction that don't exist in GA), yet it would serve its purpose (which is to facilitate comparisons with other accents.) better than ⟨ɜ⟩ because the former is the typical transcription in RP and probably also Australian English.

Let me also quote one of my previous messages: GA isn't Scottish English and writing /iː, uː, ɜː, ɑː, ɔː/ isn't wrong in any way, just somewhat redundant but it's also redundant in the case of traditional RP in which all of the tense vowels are transcribed with symbols different than those used for the lax vowels, even those that shouldn't (meaning the /ə–ɜː/ pair - the tense vowel should be written with ⟨əː⟩. If unrounding and lengthening of /ɒ/ triggered the loss of distinctive length in GA then perhaps it should be written with /ɑ/ as well - indeed, that's what Geoff Lindsey has been arguing for).

(Redacted)

I think that changing ⟨ɔ⟩ to ⟨ɒː⟩ would be an overkill and I'm dropping that proposal. Kbb2 (ex. Mr KEBAB) ( talk) 00:37, 14 March 2019 (UTC)

@ Nardog and Wolfdog: I've cleaned up the mess, hope this is more readable. If someone really wants to read the original messages, check the talk page history. Kbb2 (ex. Mr KEBAB) ( talk) 01:24, 21 March 2019 (UTC)

I can see why you might be confused. Most of my original message is about these edits to the vowel table. Apparently I thought that would be self-evident and so I omitted that piece of information. Kbb2 (ex. Mr KEBAB) ( talk) 11:08, 21 March 2019 (UTC)

I follow your length-marks argument, but it doesn't particularly convince me. As far as I seem to be able to think, Americans don't consciously hear any length distinctions -- nuff said. On the other hand, I like your ⟨ɔ⟩ → ⟨ɒ⟩ argument for purely biased personal reasons, yet still in the end agree that it's probably overkill and would upset others. As far as the Wikipedia audio goes anyway, [ʌ] sounds like the STRUT vowel, while [ɐ] aligns to the PALM/LOT vowel. As I think you're saying, Americans' /ɜːr/ really may as well be written /ər/. That one I'd wholly support. (However, if sources don't back any of this, do we have anything to stand on?) Wolfdog ( talk) 21:13, 21 March 2019 (UTC)

I think the recording of [ ʌ is bad: more central rather than back. — Eru· tuon 01:20, 23 March 2019 (UTC)

@ Wolfdog: I'm not talking about consciously hearing length distinctions but about aligning transcriptions in this article with Help:IPA/English and Received Pronunciation as we're already being redundant by transcribing NURSE with /ɜ/ (which isn't something I'm strongly opposed to, I'd rather add the length marks than get rid of ⟨ɜ⟩).

Take American /ɑ/ for instance - to me it's almost the same as RP /ɑː/ and typically doesn't have the length of /ɒ/. If you transferred GA /ɑ/ to Dutch it'd be interpreted as /aː/, not /ɑ/ because of the length (native speakers of Dutch are used to hearing back(ish) variants of /aː/ so this proves that Dutch has phonemic vowel length, so ⟨ɑ⟩ is actually a somewhat redundant transcription of what is an underlying /a/). Furthermore, GA /ɑ/ and RP /ɑː/ have pretty much the same duration as Australian and New Zealand /ɐː/ (not to mention the fact that Australian /ɐː/ and GA /ɑ/ are almost exactly the same vowel if you distinguish between PALM and THOUGHT) and these are dialects which feature an actual phonemic length distinction. We have to remember that this article doesn't exist in a vacuum and there are other articles or guides which are just as easily accessible to our readers as this one. Furthermore, there exist thousands of IPA transcriptions of American and Canadian names of places, people, etc. on Wikipedia which do use the length mark because that's how Help:IPA/English works.

If you, for example, elongated /ɪ/ or /ʊ/ before a final voiceless consonant so that it had the same length as /i/ or /u/, chances are that the final consonant would be perceived as voiced. The point is that all length distinctions in GA co-occur with quality distinctions (much like in RP and Standard German - in both there really aren't that many pairs based on length alone), and one of the atributes has to be phonologically redundant. I'm saying that it is more likely to be quality than length (and TRAP should be paired with PALM, by the way).

What I write with /i/, /u/ and /e/ in this thread are what I consider to be underlying values of KIT, FOOT and DRESS (short close front, short close back and short mid front). This audio clip sounds to me like [ˈmɛːri] (not *[ˈmɛəri]) [ˈmæri] and [ˈmɛri], with the first and third vowel being differentiated exclusively by length, as in RP (though the third vowel may be more central than RP /e/, but that's probably irrelevant). It may be one of the (relatively) few examples where length does have a phonemic value in North American English. It also seems to me that it can help to differentiate between different American dialects, because Boston /ɒː/ is not the same as NYC /ɒ/ (which may be the same thing as RP /ɒ/). The latter is a tense vowel which can occur in all positions, whereas the former is a lax vowel that cannot occur word-finally (and, in some analyses, syllable-finally). Their length also doesn't match and speakers of Boston English must shorten their /ɒː/ if they want to make their RP LOT vowel sound authentic when imitating that accent. If they won't, chances are that this /ɒː/ will be perceived as a very dark variety of /ɑː/. Until very recently, English-language vowel changes before historic /r/ equated Boston /ɒː/ with RP /ɒ/, which isn't correct (see [1] and [2]).

The additional problem with using ⟨ɒː⟩ for THOUGHT is that while [ ɒː is the actual realization in most environments, before /r/ it is open-mid [ ɔː or higher. We would have to write the merged NORTH–FORCE vowel as /oʊr/ if we wanted to show that it's closer than the main allophone of /ɔ/ (or /ɒː/, if that's how we'd write it).

The reason for which you consider our recording of [ ɐ as GA /ɑ/ might have a lot to do with context: it's a long word-final vowel. You could as well interpret it as /ʌ/ if it was short enough and if it replaced a native GA /ʌ/ in an actual English word. I have no idea whether you can do it, I certainly can't (I can barely handle Audacity and similar software). Kbb2 (ex. Mr KEBAB) ( talk) 19:31, 23 March 2019 (UTC)

@ Aeusoes1, Erutuon, LiliCharlie, Officer781, and W. P. Uzer: Any thoughts? I'd really appreciate your opinion. Nardog ( talk) 17:33, 22 March 2019 (UTC)

Split from Vowels (again)

One thought: the kit vowel being the short version of the fleece vowel doesn't really ring true for me (though I probably don't speak General American exactly). The kit vowel is usually mid and central, and it seems to be the same phoneme as schwa. It is only close before /ŋ/ or /ɡ/ (where it could also be assigned to the phoneme /i/ because even if I try to contrast ig and eeg, there isn't much of a difference). The adverb just (as opposed to the adjective just) has a schwa because it is usually unstressed, but when I stress it, I don't pronounce it like the adjective just (that is, with the strut vowel), but rather with the same vowel as gist (the kit vowel). So schwa when stressed becomes the kit vowel. On the other hand, some instances of schwa feel more like the strut vowel, like at the end of a word (though in that position the vowel could be assigned to any phoneme of strut, schwa, or kit because I think only one schwa-like vowel can occur there). I don't know if these observations of my idiolect apply to General American though. — Eru· tuon 01:36, 23 March 2019 (UTC)

@ Erutuon: I would argue that there's no such thing as word-final STRUT. AFAIK, Middle English /u/, the vowel from which /ʌ/ split centuries ago couldn't occur in that position. Any short word-final vowel that is central belongs to the COMMA set, no matter how open it is. This is supported by the fact that realizations such as [ˈkɒmɐ] also occur in dialects without the FOOT–STRUT split.

Do you consistently contrast minimal pairs such as an orthodoxy vs. unorthodoxy or a large and tidy room vs. a large untidy room? Would you consistently use vowels of the same quality in both syllables of hubbub? If the answer is no, you have the STRUT–COMMA merger (you probably use [ɐ] for the stressed vowel and sometimes the word-final one and [ə] elsewhere), not the KIT–COMMA one. I'd question anyone (no offense, I mean it neutrally) who would claim such a merger as a speaker of North American English. I've only seen it reported to occur in New Zealand. The problem with the COMMA vowel is that it is inherently unstressed and its allophonic range covers a rather large mid central area, starting from near /ɪ/ and ending near /ʌ/ or perhaps encroaching on one or both of them.

Because /ə/ is an unstressed-only vowel, anyone who has the weak vowel merger has already ceased to contrast /ə/ and /ɪ/ (unless there are any minimal pairs that could still be contrastive before velar consonants). But the weak vowel merger doesn't imply a lack of the STRUT–COMMA one. And if /ə/ and /ʌ/ belong to one phoneme, that automatically means that /ɪ/ is a different phoneme to you.

Would you say that the final vowel in happy and pedigree is exactly the same, or that the second one is longer? Kbb2 (ex. Mr KEBAB) ( talk) 19:31, 23 March 2019 (UTC)

When I say "strut vowel", I'm mainly talking about the vowel quality found in the strut set and discussing how to phonemically assign sounds without a-priori biases. I'm aware that the strut set, kit set, and comma set are defined mutually exclusively. I am questioning whether their apportionment of words is accurate for my own pronunciation.

I can contrast an orthodoxy and unorthodoxy if I try, by using a more open vowel in un-, but generally they are the same. For me the vowel of unstressed un- is simply schwa unless I specially emphasize it, in which case it's strutty. I think that doesn't disprove schwa being the same as the vowel of kit as evidenced by stressed adverbial just sounding like gist and not like adjectival just. Why can't both unstressed un- and kit have schwa?

I can't discern any contrast between the final vowels of happy and pedigree or other similar pairs. — Eru· tuon 20:02, 23 March 2019 (UTC)

I should say I'm not totally sure about how best to phonemically classify strut-like or kit-like vowels and schwa because as you say schwa has wide variation. The example of just is the only clear evidence that I've found for my kit–schwa notion. But it makes me not entirely confident in the "strut merges with schwa" idea. That idea also makes it harder to transcribe the hurry–furry distinction where it occurs; for me the most obvious transcriptions would be /ˈhəri/ and /ˈfʌri/, making ⟨ɜ⟩ phonemically (not diaphonemically) unnecessary (as it is in General American). — Eru· tuon 20:41, 23 March 2019 (UTC)

@ Erutuon: English schwa is not the only one which behaves strangely in word-final positions. In Dutch, final /ə/ is raised and rounded to [ ɵ, which is the main allophone of /ʏ/. Compare two possible pronunciations of maken ('to make'): [ˈmaːkɵ] and [ˈmaːkən] (when the second vowel is not further reduced to a mere syllabicity of the nasal), which differ similarly to English comma [-ɐ] and commas [-əz]. In German, /ə/ is consistently raised towards cardinal [ ɨ (without ever reaching it) in positions in which it is likely to contrast with /ɐ/. So the evidence that the final short central vowel in English should be considered to be a variant of /ə/ is not just historical but cross-linguistic. Also, most of the spellings of final /ə/ don't match with how /ʌ/ is normally written, and that doesn't seem to be the case with other vowels.

Further evidence that comma contains a final /ə/ is that it is only in utterance-final (and maybe in some dialects also the utterance-initial one) position where the [ɐ] allophone is used with any consistency, not just any word-final one. Try saying a few sentences with final /ə/ being followed by various sounds to see whether you consistently use this variant. I bet you don't and no-one does. This is why I reject the idea that comma is /ˈkɒmʌ/ for anyone who genuinely contrasts /ə/ with /ʌ/. If you genuinely contrast the two vowels then you consistently realize /ʌ/ as [ ɐ or [ ʌ.

If [ʌnˈɔrθədɑksi] is a mere emphatic variant to you then it's a spelling pronunciation, perhaps better transcribed [ˌʌnˈɔrθədɑksi] (just as an orthodoxy can be /ˌæn ˈɔrθədɑksi/ (or /ˈæn ˌ-/). What about and tidy vs. untidy? Is it the same thing? Is your usual pronunciation of hubbub [ˈhʌbəb] (or maybe even [ˈhʌbɨ̞b])?

I think that /ʌnˈɔːrθədɒksi/, /ʌnˈtaɪdi/ and /ˈhʌbʌb/ are the original pronunciations after the foot-strut split. How would you explain the fact the they're mostly preserved in Southern England? It's not just RP that contrasts them but they're contrasted in the whole London area and beyond (basically wherever STRUT has an [ɐ]-like realization). In Wales, for instance, the reason for their failure to contrast /ə/ with /ʌ/ is Welsh, in which only /ə/ exist and it can be stressed. In other accents it may be just an approximation of the foot-strut split as it manifests in RP. Maybe in the US the strut-comma merger is (sometimes) a regional innovation, a reduction of unstressed /ʌ/ to /ə/ (but the stressed allophone is still open). I think it's likely to be the case because in England an open [ ɐ in stressed syllables seems to imply pronunciations such as /ʌnˈɔːθədɒksi/, /ʌnˈtaɪdi/ which contrast with /ənˈɔːθədɒksi/ and /ənˈtaɪdi/.

If stressed adverbial just sounds like gist then that's just an idiolectal feature of your speech (or perhaps one of the features of your regional accent). Do you perhaps contrast Rosa's /ˈroʊzəz/ and roses /ˈroʊzɪz/? Because the weak vowel merger already makes /ə/ and /ɪ/ pretty much non-contrastive. The reason Australian English is said to feature /ɪ/ in its vowel system and NZE is said to have merged it with /ə/ is purely phonetic (you can compare it with the status of /h/ vs. /ŋ/ in most accents of English) as the two vowels don't seem to contrast in either accent (except, again, before velars - but what are the minimal pairs?). In AuE, /ɪ/ is a strong vowel of an [ i type, phonetically very different from the weak /ə/ (non-final range: [ ɨ̞ ~ ə). In NZE, both vowels are [ ɨ̞ ~ ə, with the exact same phonetic range. But in both AuE and NZE, whether we consider /ɪ/ and /ə/ to be separate vowels or not, /ʌ/ is mostly separate from the two, being an open central vowel (a short equivalent of the long open central /ɑː/).

I think Wells's idea of weak and strong vowels is great. I'd go a step further than him and use the X-SAMPA symbol for the schwa ⟨@⟩ to show that it's not exactly like other vowels (this is actually one of the reasons the NURSE vowel is written with ⟨ɜː⟩ in RP and Australian English - it's to show that it's a strong vowel and that its quality is not influenced by the neighboring sounds. It's not to show that it's open-mid, although it can be for some RP speakers). That way, you can still consider KIT and STRUT to contrast with each other and to be neutralized as /@/ when unstressed. I'd label this /@/ as a neutral close-to-open central vowel. The second vowel in roses can also be transcribed with an X-SAMPA symbol ⟨I⟩ (though in some fonts this can look like ⟨l⟩ which isn't a good thing) and I'd label it as a neutral close front-to-central vowel.

If your stressed STRUT is opener than mid then the historical evidence suggests that you have the strut-comma merger, not the kit-comma one. Kbb2 (ex. Mr KEBAB) ( talk) 13:18, 24 March 2019 (UTC)

As far as I can tell, I don't contrast Rosa's and roses or and tidy and untidy unless I'm especially emphasizing the relevant vowel. I'm not totally sure about hubbub, but I might pronounce a strut vowel in the second syllable even though it's unstressed, rather than a higher vowel as would be expected if it were a schwa in a closed syllable.

I am not sure I understand the point you're making by saying that the strut set was distinguished historically and is still distinguished in some accents. I'm aware of that and of the vowels it typically originated from in Early Modern English. Are you saying that because unstressed strut vowels became schwas in a particular accent (for instance in unorthodoxy), this establishes that the best analysis is that strut and schwa are the same phoneme? Perhaps most or all cases of unstressed strut merged with schwa, but there is still a strut phoneme that does not occur in all the same places as the strut phoneme in the dialects that preserve the full contrast but that is still different from schwa.

I suppose a slightly different analysis of the adverbial just and gist merger, taking into account what you mention about the wide variety of allophones of schwa, would be that the allophone of schwa in a closed syllable is closer and therefore it ends up being within the phonetic range of the vowel of kit rather than the vowel of strut when I stress it, and is therefore best identified with kit rather than strut. I think my idiolectal pronunciation might be used more generally because I recall people pronouncing the adjectival just like the adverbial just, with a non-strut-like vowel; it gets on my nerves because it sounds like they haven't heard the adjectival just used enough to know how to pronounce it right. — Eru· tuon 20:55, 24 March 2019 (UTC)

@ Erutuon: So your strut-comma merger might be in progress. If your hubbub fluctuates between [ˈhʌbəb] and the RP-like [ˈhʌbʌb] it can still be analyzed as having the same vowel phoneme as the second syllables of balance and comma, it's just that this /ə/ can have a rather unexpected [ʌ] quality in the second syllable of hubbub, for whatever reason. Hubbub, pick-up etc. don't form minimal pairs with other words when it comes to the /ʌ–ə/ contrast.

I'm saying that historical evidence suggest an overall much closer connection between STRUT and COMMA rather than between the latter and KIT. Thousands if not milions of Brits use [ə] for STRUT because either that's their native pronunciation, or they want to approximate the RP realization. And AmE as a whole is often analyzed as possessing only one mid-to-open central vowel /ə/. There's no need for a different analysis, no matter how central your KIT is. Traditional Cardiff English also has a pretty central /ɪ/ yet STRUT indisputably belongs to the same set as COMMA. This might not be quite the case in your accent yet, but if you don't contrast the few minimal pairs such as unorthodoxy and an orthodoxy and realize the first syllable of both with [ə], it's quite clear that you mostly merge the two vowels (you can't really speak of a phonemic contrast when there are no minimal pairs and the sounds that are supposed to contrast are phonetically so similar to each other. It's unlike the situation of /h/ and /ŋ/).

LPD lists the following pronunciations of stressed adverbial just: /dʒʌst, dʒəst, dʒest/. The unstressed forms are /dʒəst/ and /dʒɪst/, with the latter being the only non-RP form listed here. So maybe even your stressed adverbial just can be analyzed as /dʒɪst/, in which case you simply always use the strong form. Or not - see also [3]. Kbb2 (ex. Mr KEBAB) ( talk) 10:55, 25 March 2019 (UTC)

I've heard some of the dialects that have schwa-like strut, but am not sure the relevance of what other dialects of English do in relation to these vowels. It is possible that my speech is like them in this way, or not. Subjectively, a schwa-like strut (like in Welsh or Northern English) strikes me as rather different from my own speech.

Hubbub does seem to have an unstressed strut vowel; if I pronounce it with the same vowel qualities as muppet (meaning with the second vowel higher than the first) it sounds wrong. Even though that's not a minimal pair (I couldn't think of any words ending in /əb/), I wouldn't expect the surrounding consonants to trigger a different vowel height and it does seem like there's a phonemic contrast here, though as we have established not in pairs like unorthodoxy and an orthodoxy where other dialects would have a contrast. But I find that the kit vowel sometimes more closely matches the quality of a schwa, as in get it, which I'd transcribe , though you might prefer . It's variable though; sometimes I pronounce the first vowel higher. I suppose that stressed adverbial just can be understood as an isolated case that just happens to have kit, or that it has a rare stressed schwa, but it seems natural to me for schwa to transform into kit.

I agree that my accent behaves differently from New Zealand English where it sounds like strut does not merge with schwa where my accent has merged it (I wasn't specifically paying attention to unstressed strut the last time I watched a New Zealand show), but kit seems to merge with schwa. The identification of kit with schwa in my accent isn't quite as clear as in NZE. But it feels like most cases of unstressed strut have merged with schwa while perhaps a few have not, and that generally schwa is more similar to kit than to strut, even though the allophonic range of schwa probably includes both. I think my accent is somewhat idiosyncratic, but I've heard some other North Americans (probably mainly Californians and Canadians in TV) who also have a schwa-y kit vowel and might therefore display a similar pattern.

It would be neater if strut completely merged with schwa, resulting in one less phoneme, but that doesn't strike me as making sense of my own speech. It is also possible that many cases of strut merged with schwa, but then schwa developed a closer association with kit so that strut and schwa can't be identified as the same phoneme. Anyway, you don't have to believe this and it certainly isn't Wikipedia-worthy because apparently it's just me saying it. — Eru· tuon 18:43, 25 March 2019 (UTC)

The latter symbol is used to describe the result of fronting /ʌ/ in the section "Fronting of short u." Therefore, I think possibly this vowel should be the one used to describe the General American short u. I am going to make the edits. However, I do respect that anybody else has the right to revert them. LakeKayak ( talk) 21:14, 29 December 2016 (UTC)

Done. In the end only three cases of [ʌ] were altered to [ʌ̈]. Over and out. LakeKayak ( talk) 00:26, 30 December 2016 (UTC)

How is this different from [ɜ]? If it's somewhat fronted but not completely centralized, then it feels like it should be [ʌ̟] instead. Nardog ( talk) 04:01, 6 June 2017 (UTC)

@ Nardog: It was simply the notation already used on the page, and it didn't seem appropriate at the time to change everything. If you strongly feel otherwise, go ahead. I don't object. LakeKayak ( talk) 18:30, 7 June 2017 (UTC)

[ɜ] is typically the representation of the Southern STRUT vowel. I'd be wary of using that. Listen to it:

That certainly doesn't sound like a General American "uh" to me. Wolfdog ( talk) 00:00, 8 June 2017 (UTC)

Well according to Relative articulation#Centralized vowels a dieresis doesn't necessarily mean completely centralized, so I'm not implementing a change at the moment. Whatever the case, any transcription on Wikipedia must adhere to what the relevant reliable source describes rather than one's empirical and/or anecdotal evidence. Nardog ( talk) 05:19, 8 June 2017 (UTC)

Using ⟨ɜ⟩ for the STRUT vowel is a very bad idea. Nobody does that (at least in phonemic transcription) and the symbol is strongly associated with the NURSE vowel. It was only in 1993 when the phonetic value ⟨ɜ⟩ was defined as open-mid central unrounded. Before that, it was available to transcribers as a symbol alternative to ⟨ə⟩, which was defined as a central vowel with undefined height and roundedness. It could indicate a vowel that was as high as near-close and as open as near-open and it could be either rounded or unrounded. The symbol ⟨ɐ⟩ was available to use for languages with two contrastive schwas - one close (near-close or close-mid) and one open (open-mid or near-open), as in Standard German.

Wells uses ⟨ɜ⟩ (with the length mark) for the NURSE vowel to show that, unlike /ə/, /ɜ/ is a strong vowel (at least in RP, Australian English, etc.), which means that it can occur in stressed and unstressed syllables alike. That, however, is not the case in GA, in which /ɜ/ cannot be said to contrast with either /ə/ or /ʌ/. The most logical analyses of words like nurse or word are /nərs, wərd/ or /nʌrs, wʌrd/. We write those words as /nɜrs, wɜrd/ to facilitate comparisons with other accents, following the AoE. Kbb2 (ex. Mr KEBAB) ( talk) 11:50, 4 April 2019 (UTC)

I notice Kbb2 has a tendency to resist and change narrow transcriptions to broader ones. I get the thinking behind this (that we go with a broad phonetic transcription), but, especially when we have exact audio files of speakers, why can't we have very narrow transcriptions? (Maybe we could represent both?) Wolfdog ( talk) 12:34, 13 April 2019 (UTC)

@ Wolfdog: I assume you're talking about this revert? I reverted you because /l/ in the syllable coda before a consonant or a pause is always velarized (or vocalized) in Standard English (save for very few speakers). The canonical value of the IPA sign ⟨l⟩ does signify a lateral approximant without a secondary articulation, but it can be apical or laminal and dental, alveolar or postalveolar, depending on the language. I see no reason to mark the velarized allophone of /l/ because of how predictable its occurrence is.

The vowel in [mɛlk] probably also isn't a perfect cardinal [ɛ], no? What about the velar stop? Can it be preglottalized or perhaps slightly aspirated? If so, that's not a perfect canonical [k] either.

Are there any other edits of mine you'd like to discuss? Kbb2 (ex. Mr KEBAB) ( talk) 13:02, 13 April 2019 (UTC)

The reason to mark the velarized allophone is because it's pronounced that way in the exact audio, which is also a great example of how Americans in general tend to pronounce /l/. I don't understand the urge to change that. I've just noticed you have a tendency to remove narrow markings that make sense and could be useful. Why can't [ˈpɑɹɾi], for instance, be [ˈpʰɑɹɾi]... which is actually what a listener will hear in the audio? Wolfdog ( talk) 13:45, 13 April 2019 (UTC)

@ Wolfdog: Ok, but what about [ɛ] and [k]?

I don't find it a particularly great example of that. Coda /l/ is velarized by almost all speakers of what you could call Standard English. It's not a marked feature of any accent (in fact, the opposite of that is a marked feature of Welsh and Irish English) and it's allophonic. Plus, you're still assuming that [l] can only denote a 'neutral' lateral approximant without a secondary articulation. It's not true. Besides, why should we only denote the velarization and not the fact that it's apical alveolar? Because the canonical value of IPA ⟨l⟩ can be apical or laminal and dental, alveolar or postalveolar. It's not apical alveolar by definition. Plus, [ɫ] (assuming that this denotes a velarized apical alveolar [l]) isn't the only allophone of /l/, as this consonant can be advanced to the dental position before /θ/ and retracted to the postalveolar place of articulation before /r/. It's also devoiced after onset /p, t, k/.

I write the lateral approximant in [mɛlk] as such because I consider the velarization diacritic to be implied by context, just as the apical diacritic is also implied by context here. In play [pleɪ], the velarized, apical and devoiced diacritics are implied by context as well, whereas in wealth [wɛlθ] the implied diacritics are velarized, laminal and dental. We follow this practice on Australian English phonology, English phonology, Received Pronunciation and other articles. There's no need for a special treatment of a few articles to the exclusion of others.

Similarly, the initial /p/ in party is aspirated by almost all speakers of Standard English. It's the standard allophone of word-initial /p/ and using an actual unaspirated [p] sounds like /b/ to native speakers. You need to differentiate between the two transcriptions. The only signs in [ˈpɑɹɾi] that you could read literally are the first one (the primary stress mark) and the fifth one (the symbol for an alveolar tap). ⟨ɑ⟩ doesn't denote a cardinal [ ɑ but a sound that is more front than that and ⟨i⟩ doesn't denote a cardinal [ i but a sound that is more mid-centralized than that. ⟨ɹ⟩ doesn't denote an alveolar approximant either but a postalveolar or a retroflex one, so ⟨p⟩ in this context denotes an aspirated voiceless bilabial plosive. Broad [ˈpɑɹɾi] translates into narrow [ˈpʰɑ̟ɹ̠ɾi̽].

The voiced plosives are in fact typically unaspirated voiceless or partially voiced, so writing [ˈpʰɑɹɾi] for party and [du] (rather than [d̥u]) for do would be inconsistent even when you disregard the other inconsistencies.

This isn't the first time I see an editor say that using the dark l symbol or the aspiration diacritic somehow makes the transcription narrow. We could write milk with the dark l symbol if we talked about velarization. If we wanted to be consistent, we could use narrow transcription in all places - but there's no point in doing that. Kbb2 (ex. Mr KEBAB) ( talk) 14:29, 13 April 2019 (UTC)

You're turning back transcriptions where you feel extra detail is already implied by context, but my argument is let's not assume anything is implied. Why make assumptions when we can be straightforward to readers at all levels of familiarity with IPA? This is a website that anyone can view and edit, not a linguistics library. As for your party argument, I get that you're worried about a slippery slope, but [pʰ] is a general established feature of American English (and even English overall, as you say -- and it is clearly what the speaker is doing in this one particular recording) in the exact same way as [ɫ] -- these are in fact the standard AmE allophones -- whereas [i̽] is a token that you're hearing, more open to subjective different transcriptions based on listener. So, yes, I get that initial /p/ in party is aspirated by almost all speakers of Standard English, yet I'm not getting why that means we should leave out the aspiration diacritic. So what? You think aspiration is assumed so let's leave it out, and I'm saying let's not assume that. Not everyone here is an IPA expert: let's leave it in.

I also think you're missing that I'm saying when we have exact audio recordings, even [ˈpʰɑ̟ɹ̠ɾi] would be perfectly appropriate and acceptable. I never brought up the Western New England English example you keep arguing about, but I guess I will now. That revert just continued to make me aware of how you revert narrowing transcriptions that can be useful: WNEE speakers certainly use a velarized /l/ and I still don't really see how it could hurt to show that. I wonder what other editors think. Wolfdog ( talk) 23:31, 13 April 2019 (UTC)

I lean towards more narrow transcriptions, even when the phonetic detail conveys predictable allophony, but I do understand the concern about too much detail. Perhaps a good middle ground here would be, outside of immediate discussions of a particular feature (e.g. vowel nasalization, precise vowel height, de-voicedness of lenis stops), we stick to transcribing English with the level of detail we would give it in an IPA for X convention if it were another language we were transcribing for Wikipedia articles. That is, we avoid the level of detail that requires diacritics. Off the top of my head, I imagine that we'd want to indicate predictable vowel length, aspiration, l-velarization, vowel reduction, flapping, and glottalization. I'm sure there's more we could indicate, but we'd want to come to an agreement on it. We can probably get away with avoiding r-coloring of vowels and syllabic vowels consonants, but we'd want to be consistent on how we represent those. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 00:29, 14 April 2019 (UTC)

@ Wolfdog: "Let's not assume anything is implied" is never how the IPA works. All sorts of stuff is implied in most transcriptions; in fact, this is how you transcribe words yourself whenever you don't use fully narrow transcription. This is why I asked whether [ɛ] in milk is a cardinal [ɛ] and whether the final [k] was an actual [k] or a preglottalized [ʔk]. It was a recent revert of yours, so it made sense to me to bring it up (especially since it's relevant to the topic of this discussion). You should read the quotes on User:Nardog.

The voiceless stops /p, t, k/ have many allophones in English, and generally they're most variable in the word-final position. How would you transcribe the word top? [tʰɑp]? If so, that's inconsistent. If you wrote it [tʰɑp̚] that'd be only slightly closer to the most common variant as the citation form varies between [tʰɑp] (perhaps with slight aspiration of the final /p/) and [tʰɑp̚], but when top occurs immediately before a word that begins with a vowel it can no longer be pronounced [tʰɑp̚] but either [tʰɑp] or [tʰɑp͡ʔ] (the final /p/ may be glottaly masked). [tʰɑp͡ʔ] may sound the same as [tʰɑp̚] but it's physically impossible to pronounce an intervocalic unreleased plosive. So, given the fact that final /p, t, k/ are actually more variable than initial /p, t, k/ it only makes sense to use the broad transcription of ⟨p, t, k⟩ for those sounds, so what's the problem of transcribing initial /p, t, k/ with ⟨p, t, k⟩? They're invariably aspirated, so [pʰ, tʰ, kʰ] are the only possible allophones.

Narrow transcription of English plosives is just as problematic in the case of RP: the citation form of top is any of the following: [tʰɒp, tʰɒpʰ, tʰɒʔp] (or with affrication instead of aspiration, the vowel is also variable [ɑ ~ ɒ ~ ɒ̝ ~ ɔ̞]), but before a vowel only [tʰɒp, tʰɒpʰ] are possible. So what do we do? We write it [tɒp]. I think we should strive for more or less the same level of narrowness across all articles on English accents and to leave our excessive detail that we probably wouldn't include in articles on the pronunciation of other languages.

Note that speakers who consistently release their final stops don't sound native in American English, so [tʰɑp] can't be said to be a narrow transcription of top.

You're worried that laymen may assume ⟨l⟩ to mean a clear lateral approximant - but that's on them, not on us (or perhaps it's also on us if Wikipedia is actually failing to convey that there are levels of narrowness of phonetic transcription). American /l/ is consistently velarized in all positions, at least in General American, so our readers can always assume [l] to be equal to [lˠ]. In fact, all allophones of /l/ in GA can be defined as velarized lateral approximants, because GA /l/ can be apical or laminal and dental, alveolar or postalveolar, depending on the environment (even though the canonical value of ⟨l⟩ in GA is a velarized apical alveolar lateral approximant).

If we were to also use the ⟨ɫ⟩ in Received Pronunciation, that'd introduce complication to transcription as RP /l/ is dark only before consonants (but not /j/ - another complication) and pauses, so that law, ceiling, million and call up would have to be transcribed with ⟨l⟩ but call and pills would be written with ⟨ɫ⟩. It's better to write [lɔː, siːlɪŋ, mɪljən, kɔːl ʌp, kɔːl, pɪlz] and let the reader figure out which laterals are velarized and which aren't (the rule is very simple, and the fact that General American /l/ is pretty much categorically velarized is even easier to remember). In General South African English call up is [kɔːɫ ʌp] but other words are pronounced the same as in RP. It's simpler to just use ⟨l⟩ for these sounds. Also, younger speakers of RP (but not of General South African English) may vocalize the final /l/ in call and pills, so they should actually be written [k̠ʰɔ̝ːɫ ~ k̠ʰɔ̝ːʊ̯] and [pʰɪ̈ɫz̥ ~ pʰɪ̈ɤ̯z̥] in truly narrow transcription (the /l/ in call up can also be [ɫ] or [ʊ̯] if you put a glottal stop in front of up. So even in RP there's some variability).

It's not our goal to transcribe the recordings but the variety of English in question (I remember that you even argued for transcribing a cot-caught merged pronunciation of water as such in this article). I don't consider [ˈpʰɑ̟ɹ̠ɾi̽] to be an acceptable transcription of party in this context, and by saying that you do you're showing that you need to better differentiate between narrow and broad variants of phonetic transcription. As you said yourself, this isn't a linguistic paper but an encyclopedia.

Also, note that IPA ⟨ɫ⟩ denotes a velarized or pharyngealized lateral (depending on the language), not just a velarized one. American /l/ is rarely (if ever) pharyngeal, so [mɛɫk] isn't really more narrow than [mɛlk], at least in some sense. Kbb2 (ex. Mr KEBAB) ( talk) 15:18, 15 April 2019 (UTC)

Thanks -- Nardog's page is helpful in some ways, though I have to read your above explanation about a million more times to wrap my head around it. I still find it odd that predictable sounds are the ones most left out rather than the ones that can most easily be left in. My study of IPA was that it was indeed largely phonetic and less so that single symbols represent in fact vast allophonic possibilities in phonetic transcription. Anyway, it just seems to be you and me mostly. If people are ever interested in coming up with a standard for American English here on Wikipedia, I'd be happy to help engage in that work. Wolfdog ( talk) 15:46, 15 April 2019 (UTC)

I for one would be in complete support of Aeusoes1's suggestion for a compromise. In fact, I really was trying to argue for some kind of middle ground too, since I recognize Kbb2 is concerned about a slippery slope. So! As Aeusoes1 said (I think more clearly than I was able to), I'd agree with indicating down to a "predictable" level of detail, and I'd be totally on-board for the specifics suggested: predictable vowel length, aspiration, l-velarization, vowel reduction, flapping, and glottalization. This would cover of course the two features I mentioned above, but not such insane level of detail as the [i̽] token that worried Kbb2. Wolfdog ( talk) 11:11, 14 April 2019 (UTC)

Side-note: I just noticed/remembered that we already had a discussion in the same vein as this with one Erutuon, Nardog etc. over a year ago titled "Aspiration". Although I just skimmed it (and will need to reread it a couple of times), it did seem there was a general consensus for a middle-ground position -- as Aeusoes1 says, now we'd want to come to an agreement on the details. Kbb2, under "Mr KEBAB", it seemed even you were more shifting towards a compromise position at that time. Erutuon seemed in favor of [pʰ] for both pay and play, whereas you commented I think using [CʰV] with vowels and [CC̥] with consonantal sonorants is a good and necessary compromise. Anyway, I'd love to participate in a discussion on the details we'd want to include. Wolfdog ( talk) 11:37, 14 April 2019 (UTC)

@ Aeusoes1: Our IPA for X conventions don't include aspiration and velarization when those aren't phonemic. Flapping is different than that because it's less predictable than aspiration (which in fact is fully predictable, as is velarization), so we should include it. I'd argue that using distinctively fully voiced stops in the citation forms of words such as bad (especially in the case of the final /d/) sounds just as non-native as not aspirating initial /p, t, k/, even though the vowel length is sufficient to differentiate between bad and bat.

By predictable vowel length do you mean simply transcribing the free vowels with ⟨ː⟩, including in phonemic transcription (so that keep is transcribed /kiːp/ and father /ˈfɑːðər/? If so, I fully agree with that.

What exactly do you mean by vowel reduction? /ə/ is widely considered to have phonemic status in English and that's how we treat it here.

I think that NURSE and LETTER should be transcribed as r-colored. Wells (or whoever it was) made an unfortunate decision to treat those as vowels, rather than as syllabic consonantal approximants. Transcriptions such as [ɜːɹ] or [əɹ] are, IMO, bad practice and even LPD and CEPD write those as r-colored, even though their transcriptions are otherwise to a large extent phonemic. The only difference between SQUARE and NURSE/LETTER is that the latter is a syllabic consonant (sometimes pronounced with a slight schwa onglide), whereas the latter is a front vowel followed by a consonant. Transcriptions such as [ɜːɹ] or [əɹ] fail to show that.

If we want to show glottalization then I think that we also want to explicitly denote syllabic consonants, because pronunciations such as [ˈbʌʔən] for button are probably still used by a minority of speakers. Plus, it's a widely accepted practice to explicitly denote optional schwas (or syllabic sonorants, same thing really) in transcription. Kbb2 (ex. Mr KEBAB) ( talk) 15:18, 15 April 2019 (UTC)

@ Kbb2: Outside of aspiration and velarization, I think we're in broad agreement on most of what you're saying. I think diacritics are the better framework to consider than what would sound non-native or unnatural if it were missing. We should be phonetically precise enough to be accurate, but not so much that we would need combining diacritics on vowels.

I think that indicating aspiration and velarization would be helpful to lay readers. We specialists consider the distribution of aspirated phones to be predictable, but lay readers may not (even if they're native English speakers who produce it). Indicating aspiration is more phonetically accurate and doesn't fall down the trap of too much phonetic detail that we want to avoid.

Velarization is only predictable within a given dialect, and a reader shouldn't be expected to know it. I think we wouldn't want to indicate velarization only if we don't think we could provide consistent accuracy in transcriptions. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 21:15, 15 April 2019 (UTC)

I, of course, still agree will Aeusoes1's points. I am and have been concerned with lay readers. As long as we come to some common agreement, we can practically transcribe American English however we want within IPA parameters. Wolfdog ( talk) 13:45, 16 April 2019 (UTC)

Ok, so we add aspiration, length marks, remove the rhotacized diacritic and write NURSE with [ɜːɹ] and LETTER with [əɹ] and transcribe the dark l's. Would everyone be happy with that? Kbb2 (ex. Mr KEBAB) ( talk) 08:20, 18 April 2019 (UTC)

That all sounds good to me... I think, ha. Can I just get three clarifications, Kbb2? See below:

1. We're talking about coming to an agreement on what basic level of narrow transcription to use across all articles falling under American English, correct?
2. Can you remind me what we're meaning here by predictability of length marks (which is non-phonemic but still fairly predictable in AmE)? i.e., Are we talking about, for example, where lengthening tends to happen in AmE before voiced consonants but not before voiceless ones? That use of length marks? In the articles you've edited so far, I think you've been simply aligning the length marks to the BrE style; is there a reason to do this for AmE?
3. Kbb2, I thought you would prefer something like [ɚ] to represent NURSE and lettER rather than [ɜːɹ] and [əɹ], no? (Though I do love your use of the inverted "R": [ɹ].)

Wolfdog ( talk) 12:21, 18 April 2019 (UTC)

I am in favor of showing aspiration and velarization, but I second Wolfdog's request for clarification about length marks, and prefer transcribing both nurse and letter as an r-colored schwa [ɚ]. — Eru· tuon 19:50, 18 April 2019 (UTC)

Kbb2, can you clarify? Wolfdog ( talk) 19:06, 24 April 2019 (UTC)

OK, so I propose GenAm transcription should include these features (some of which are being newly proposed, I think):

• /l/ is transcribed in all environments as [ɫ]
• /r/ is transcribed in all environments as [ɹ]
• Both NURSE and lettER are transcribed in all environments as [əɹ]
• Aspiration is shown in appropriate environments; e.g., popping as [ˈpʰɑpɪŋ]
• /æ/ is transcribed before /m/ and /n/ as [ɛə]
• /t/ is transcribed in appropriate, highly predictable environments as [ʔ]; e.g. catfish as [kʰæʔfɪʃ]
• /t/ and /d/ are transcribed in appropriate, highly predictable environments as [ɾ]; e.g. rating [ˈɹeɪɾɪŋ] (though some words, like skeletal, appear to vary more greatly: [ˈskɛɫətʰəɫ] or [ˈskɛɫəɾəɫ])

There may be others features to delineate, like the inclusion or exclusion of length marks; I personally feel this is a minor element of AmE, trivial for GenAm, but we can discuss. Thanks, all. Wolfdog ( talk) 20:10, 30 April 2019 (UTC)

GenAm has dark l in all contexts?! — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 20:08, 1 May 2019 (UTC)

Have you seen the bullet on L-velarization under "Consonants"? Wolfdog ( talk) 13:06, 3 May 2019 (UTC)

I see what you mean. That looks very suspect and runs contrary to what I've seen in general about GA. I'll have to dig into the matter, but based on what I know so far, I wouldn't be in favor of velarizing l in all contexts, just in the syllable coda. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 19:35, 3 May 2019 (UTC)

It's a prominent feature of Received Pronunciation (and I guess many accents in England) that l is noticeably less velarized in certain positions than in General American or Australian. It's one feature that I have occasionally noticed in British actors who are using an American accent: they failed to sufficiently velarize their syllable-initial l, giving an odd sort of Irish quality to certain words (sometimes mixed with incongruous features like r-dropping). In my own speech the main contrast between initial and final l seems to be that the final l is somewhat more velarized and the tongue often doesn't touch the teeth or alveolar ridge (l-vocalization). — Eru· tuon 21:41, 3 May 2019 (UTC)

I agree with every one of Erutuon's above sentiments and have made similar observations myself. There may be slight articulatory differences between initial and final /l/ in AmE, but they're both still fairly velar. John C. Wells' Accents of English (Volume 3) says "GenerAm /l/ tends to be rather dark. Before stressed vowels it is neutral or only slightly velarized; preconsonantally and finally definitely dark (velarized)" (490). So... I'd still err on the side of showing velarization in all cases for our broad GenAm transcription. Certainly, as Erutuon suggests, if someone pronounced a perfectly clear initial /l/ and a perfectly dark initial /l/, only the first could sound to my ears like a non-American (or perhaps Spanish-influenced American) dialect. Even the darkest pronunciation of an initial /l/ would fly completely under my radar. Wolfdog ( talk) 13:40, 5 May 2019 (UTC)

Yeah, it seems like the onset-coda allophony present in RP is echoed in GA, except that onset /l/ can have a slight degree of velarization. But that varies from speaker to speaker. Transcribing all of these as dark would make our transcriptions more opaque by eliminating any representation of this allophony and implying that GA is like Scottish English, which has noticeably dark /l/s in all position. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 14:35, 5 May 2019 (UTC)

Well, I guess we just disagree then. Wells is saying that velarization is basically everywhere in GenAm and only differs from one context to another by degree. I think the "echoing" or light/dark allophonic distinction in GenAm is insignificant (probably even absent in some AmE speakers... I certainly don't hear it in my own conscious speech). Every American /l/ is dark compared with RP, as confirmed by Erutuon when noticing some British actors' /l/ pronunciations still sounding wrong when attempting American accents. What Erutuon is presumably noticing is a clear /l/ in non-coda positions when listening to these actors. Americans don't have any clear /l/. Wolfdog ( talk) 16:40, 5 May 2019 (UTC)

The quote you've provided indicates that Wells believes it to be "slight velarization." The difference between slight velarization and velarization is not insignificant and Wells demonstrates he sees these two as different elsewhere in Accents of English, when he says (p. 74):

"In RP, GenAm, and many other accents two perceptibly different allophones may be distinguished, clear and dark. The details of the environment in which the dark allophone is used vary, however: thus intervocalically, as in silly, RP uses clear /l/, thus [ˈsɪlɪ], while GenAm uses dark, thus [ˈsɪɫɪ]. In this word, as in others where /l/ is intervocalic (valley, yellow, column), RP thus sides with Irish English against GenAm; compare words such as belt, milk, halt, where RP and GenAm agree in having dark /l/, but Irish English uses a clear variety."

The use of a dark l in additional phonetic contexts may help account for Erutuon's perception of noticeably darker l's in American English that are clear in RP. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 17:23, 5 May 2019 (UTC)

Would you be in favor then of transcribing GenAm lilly as [ˈlɪɫi]? Wolfdog ( talk) 18:04, 5 May 2019 (UTC)

Yes I would. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 23:25, 5 May 2019 (UTC)

Do we know how this allophony for /l/ works postconsonantally in GenAm? For instance, is the /l/ in /fl/ clearer than the /l/ in /gl/ due to voiced/voiceless reasons? Wolfdog ( talk) 16:51, 6 May 2019 (UTC)

Probably. My guess is that the differences are slight and the general tendency is how Wells describes it. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 03:14, 7 May 2019 (UTC)

Here's another source that suggests "General American /l/ is velarized in all positions" (though I admit the sentence doesn't seem to make grammatical sense to me). And this source says (see under the section "Velarized alveolar lateral) that a velarized /l/ is used "in most Northern varieties of American English" which I take to mean non-Southern (including GenAm) varieties. Another source confirms (per my allophonic distinction question) that American dark /l/ appears strongest in both postvocalic and intervocalic positions, so it works for me too that we represent it that way. Wolfdog ( talk) 14:41, 7 May 2019 (UTC)

I'm interested in the Kbb2's proposal to change the transcription of the NURSE vowel to a syllabic consonant, presumably [ɹ̥]. Does anyone know of a published transcription that does this? — Marquetry28 ( talk) 06:53, 3 May 2019 (UTC)

I don't have a strong preference in any direction, but I notice sources that use [ɚː] here or [ɚ] here. I do see one use of syllabic [ɹ] here, which happens to be discussing Canadian English but which could just as easily be discussing GenAm, I suppose. Wolfdog ( talk) 14:41, 7 May 2019 (UTC)

When I first read this article, I found two vowel charts. The first was a graph of vowels for GA speakers without the cot-caught merger; the second, with. Apparently the one for speakers with the cot-caught merger was deleted, along with the reference to the source from which it came. Why was it removed? AnUnnamedUser ( talk) 02:24, 21 July 2019 (UTC) AnUnnamedUser ( talk) 02:24, 21 July 2019 (UTC)

@ AnUnnamedUser: (We're talking about this chart, which originally was uploaded as [4]) It's not useful. The article already explains that the merger is a possibility and how it works. No vowel chart on Wikipedia (maybe with one or two exceptions) covers all of the allophones of any given phoneme.

This chart, which is how the source used to depict (it's been deleted) the vowels of AmE, isn't 100% correct: /u/ is too back, /ɪ/ too front, /ɛ/ too high and [ɚ] could be too low (I think it should be close-mid, like Australian NURSE). Whether we choose this one or the Wells's one would be arbitrary, except for the fact that the latter is based on a more reliable source that actually exists.

Besides, one of your edit summaries reads Kbb2, I think you accidentally deleted the reference for the proper chart with the vowels for GenAm with the low back merger instead of the one without it. Why would I remove Wells's chart and what's "improper" about it? I've demonstrated that the monophthong chart from that Australian source is just as flawed if not more so.

The source itself has deleted the chart with the cot-caught merger, along with the New Zealand English charts and the RP ones. The RP ones are roughly the same as the ones in Received Pronunciation. The NZE ones are clearly unreliable (/e/ is too low, /ɵː/ is too close to /ʉː/, the ending point of /æʊ/ is wrong and so is the starting point of /ɐʉ/), and so is the diphthong chart of cot-caught-merged AmE (the starting point of /eɪ/ is too low, the starting points of /aɪ/ and /aʊ/ too high and the starting point of /oʊ/ is probably too low and/or not central enough (it should be [ɵʊ] or something like that). Wells's diphthong chart looks more correct and less chaotic. Kbb2 (ex. Mr KEBAB) ( talk) 06:39, 21 July 2019 (UTC)