Styles of English Phonemic Transcription

[Travis B.]

I have noticed on here that there have been a wide range of ways of transcribing English phonemes. There appear to be two extremes, attempting to transcribe English phonemes universally, trying to use one set of idealized English phonemes for all English varieties, and attempting to transcribe English phonemes dialectally, using phonemic transcription tailored specifically to the phonology of a particular English variety being spoken of. However, I have seen very many people use phonemes that appear to be somewhere in between the two, particularly applying English phonemic transcription in a more universal way while having actual transcribed phonemes that are clearly influenced by the phonology of some particular English varietie(s).

Personally, I tend to favor using one extreme or the other as far as possible, depending on context. When speaking about my own dialect, I tend to favor using dialectal phonemes that specifically fit its phonology. When speaking about any other English variety or when speaking about English phonology in a more general or abstract fashion, I tend to favor using universal phonemes that attempt to fit all English varieties, or at least the vast majority of them (considering I transcribe them with some mergers that have happened in almost all extant English varieties but which are not complete in truly all English varieties). I try my very hardest to keep the two from contaminating one another, e.g. I always mark historical phonemic vowel length and favor idealized historical phoneme qualities when transcribing universally and I refuse to allow other varieties' phonologies to influence the phonemes I use when transcribing dialectally.

So how do you guys out there transcribe English, particularly whether you are transcribing one's own variety of English or you are approaching English phonemes from a universal perspective, and why?

[Travis B.]

Somehow I should have guessed that this was going to be an uninteresting topic to others here...

[zompist]

The last time the subject came up, I provided a sound sample and invited people to transcribe it. The results were interesting: the transcriptions were close, but often disagreed, though never by more than one phonetic step. This accords with my experience researching numbers: transcriptions, even by trained linguists, are just not accurate enough to warrant, say, the level of detail you like to provide.

It's also relevant, I think, to look at how Labov approaches the problem in tracking ongoing sound change. He's looking at tiny changes and needs extreme precision— so he does not use transcriptions at all (except informally, to describe what's going on). He makes recordings and looks at the formants. So for vowel height, that gives him a scale of ~ 500 distinctions, rather than the ~ 8 given by IPA plus diacritics.

For phonemes, Labov uses an idealized set that he can use for all English dialects... I can't find where he gets it, but FWIW here are the vowels:

short vowels: i u e ʌ æ o

long upgliding: iy ey oy ay uw ow aw

long ingliding: æh ah oh

You might check out his comments in Principles of Linguistic Change: Social Factors p. 124ff.

A lot depends on what you want to use the notation for. If you were writing a reference grammar of Milwaukee English, you might want a set of phonemes that relate closely to the actual phonetics. Labov does a lot of comparative work, so it's most useful for him to have an idealized system. When I was doing my analysis of English spelling, I used a system describing my own dialect, but with symbols chosen to explicate the orthography.

[Travis B.]

Too bad I do not have Labov's book on hand, but I have seen his system before on the net, in a form similar to (but not exactly identical to) that you mention here. It is pretty similar to the system I use when trying to speak of English dialects universally, even though the exact notation I use is rather different (probably because I try to make full use of IPA and approximate idealized historical qualities, while his system as you mention it seems to be Americanist-inspired while trying to use as few non-ASCII characters as possible). Also, his system seems to be meant for North American English in particular, while I try to cover most English varieties (except those lacking the pane-pain merger, the toe-tow merger, or the meet-meat merger).

Hence when speaking universally I try to use the following vowels:

/æ ɛ eɪ̯ ɪ iː ə ɑː ɒ ʌ ɔː oʊ̯ ʊ uː aɪ̯ aʊ̯ ɔɪ̯/

and the following vowels followed by /r/:

/ær ɛr eːr ɪr iːr ɜr* ər ɑːr ɒr ʌr ɔːr** oːr** ʊr/

* To be truly universal, I would split this up into /ɛr/, /ɪr/, and /ʊr/, but this is only necessary when speaking of Scottish English and Irish English varieties.

** In practice I tend to represent these two as /ɔːr/, as normally I am not speaking of varieties which preserve this distinction.

I do not represent an /æː/ as described by Labov, as I treat that as a limited vowel split that has shown up in some North American English varieties, many Australian English varieties, and some English English varieties, just like I do not represent a split of /aɪ̯/ into /aɪ̯/ and /əɪ̯/ despite that occurring in parts of North American English.

When speaking of my own dialect, I use a rather different system:

/ɛ e ɪ i a ɜ ə ɒ ʌ o ʊ u ae̯ əe̯ ɑɔ̯ ɔɪ̯/

and the following vowels followed by /r/:

/er ir ar ɜr*** ər ʌr or ur ae̯r ɑɔ̯r ɔɪ̯r/

*** This is rare, and only shows up in a few words where unpredictable elisions have occurred that bring /ɜ/ and /r/ together.

This system does not line up exactly with the universal system, as it takes all the phonological changes that have occurred between the idealized point in time represented by the universal system and the present in my own dialect into account. For instance, /ae̯r ɑɔ̯r ɔɪ̯r/ correspond to universal /aɪ̯ər aʊ̯ər ɔɪ̯ər/, as synchronically it makes more sense to consider the "/ə/" to be epenthetic and, in fast speech, optional, rather than to consider it as being an actual phoneme.

Likewise, I have number of choices have been made in the actual representation that are based upon my own dialect's particular phonology that bring it out of alignment with the universal system. For instance, universal /æ ɛ/ becoming /ɛ ɜ/ is based purely upon that the distinction is no longer a height distinction but rather a frontness distinction (there is actually a paper on this, too bad I cannot recall enough on it to cite it). Similarly, how I map vowels before /r/ is based purely upon what makes sense synchronically with regard to vowel quality rather than attempting to match how things are mapped in the universal system. For instance, the universal /ɔːr/~/oːr/ is simply mapped to /or/, as the vowel phoneme with the closest quality happens to be /o/, instead of creating a dedicated vowel phoneme-rhotic pair "/ɔr/" to match the universal system.

[Travis B.]

What I was really interested in here, though, is how other people choose to transcribe things, and their rationales for why they do so. In particular, I have noticed a lot of people transcribing general English phonemes, corresponding to what I refer to as "universal" above, with phonemes different from the ones I use; e.g. transcribing them without marking historical phonemic vowel length, using /e o/ instead of /eɪ̯ oʊ̯/, using /əʊ̯/ instead of /oʊ̯/, and so on.

[clawgrip]

Because a lot of the time, my analysis of English is related to EFL, I tend to try to simplify my analyses as much as possible, aiming primarily to distinguish between the core elements of what I want to teach. I will avoid marking historical long vowels since it is not relevant to pronunciation. I feel that all historical long and short vowels and diphthongs are long, i.e. stressable vowels, while only /ɚ/ /ə/ /Vi/ /Vo/ are short, i.e. unstressable vowels. I still haven't worked out a way to teach it to be people that I feel will contribute to productive learning though.

[Travis B.]

That distinction I have usually seen expressed as being full versus reduced vowels, without actually invoking vowel length per se.

[clawgrip]

Might be a good way to explain the distinction. I prefer to avoid the words 'long' and 'short' since most of my students are Japanese and may assume a parallel with Japanese vowels, which has actual phonemic length.

[Travis B.]

The usual analysis I have seen is that English typically has only two reduced vowel phonemes /ə/ and unstressed /ɪ/ (realized as [ɨ]), aside from some dialects having a reduced counterpart to /oʊ̯/ (realized as [ɵ]), with /ər/ (or non-rhotically just /ə/) for what you analyze as /ɚ/, and with diphthong offglides being part of full diphthong phonemes rather than being treated as reduced themselves.

However, if one is trying to teach Japanese students, there may very well be reason to treat diphthong offglides as separate segments and to treat /ər/ as a single vowel phoneme /ɚ/, considering what the typical Japanese student is likely to be used to natively, even if it is not how one would typically analyze English varieties in a non-EFL context.

[Nortaneous]

Full: /æ ɑ ɛ ʌ ɔ e o ɪ ʊ i u/
Diphthongs: /ai au oi/ and sometimes /əi/
Vr sequences: /ɑr~ar ɚ er ir or/ and occasionally /jur/ for the words that vary between [jɚ] and [jor]
Reduced: /ə ɨ ɚ/

Standard practice, except I use /ɑ/ instead of /a/ for LOT to avoid confusion between my phonemic and phonetic transcriptions -- [a] is an allophone of /ai/ IMD. (I guess I should write /æu/ instead of /au/, but that's ugly and not terribly necessary.) Length is too much of a clusterfuck for me to bother writing; if I write it phonemically, I'd have to write it phonetically, and phonetically there are at least three different lengths that would have to be distinguished, not to mention that I have absolutely no idea whether /æ/ is short or long.

I use /ɨ/ because I generally can't be bothered to mark stress, so I'd prefer there not to be an overlap between reduced and unreduced vowels; I should also use that rhotacized 3 thing for the unreduced rhotic vowel, but the X-SAMPA -> IPA converter doesn't support that for some reason.

I haven't yet had any reason to make my transcriptions more universal than just GA.

For consonants, I write /ɾ/ as phonemic, because doing otherwise would take more effort than I can be bothered to put in.

[AnTeallach]

What I was really interested in here, though, is how other people choose to transcribe things, and their rationales for why they do so. In particular, I have noticed a lot of people transcribing general English phonemes, corresponding to what I refer to as "universal" above, with phonemes different from the ones I use; e.g. transcribing them without marking historical phonemic vowel length, using /e o/ instead of /eɪ̯ oʊ̯/, using /əʊ̯/ instead of /oʊ̯/, and so on.

There isn't really a standard way of doing it; the nearest to one is probably Wikipedia's. I don't really think it's very important which you use, as long as it's not too confusing (but I have to confess a dislike of those non-syllabicity diacritics: are they really necessary in the context?). Personally I prefer to use Wells's lexical sets rather than IPA symbols in most cases.

For my own dialect I tend to use a hybrid of the Gimson and Upton schemes for RP (using Upton's choices for TRAP, DRESS and SQUARE but Gimson's for NURSE and PRICE, so /a/ /ɛ/ /ɛː/ /ɜː/ /aɪ/), with adaptations for ways in which my speech differs from RP.

[Travis B.]

When describing present-day* General American I personally prefer to just use a subset of said universal system, even though I am not certain how well vowel length really survived into GA (considering that the father-bother merger has been posited as the result of two steps, first /ɒ/ > [ɑ], as also seen in some English English varieties, and then the neutralization of the length distinction between that and /ɑː/)

* That is, including the Mary-merry-marry merger, the mirror-nearer merger, the hurry-furry merger, and the horse-hoarse merger.

Hence when speaking of present-day GA I try to use the following vowels:

/æ ɛ eɪ̯ ɪ iː ə** ɑː ʌ** ɔː oʊ̯ ʊ uː aɪ̯ aʊ̯ ɔɪ̯/

** These are really the same phoneme, being distinguished only by stress,

and the following vowels followed by /r/:

/ɛr ɪr ɜr*** ər*** ɑːr ɔːr ʊr/

*** These are also really the same phoneme pair, being distinguished only by stress.

I sometimes wonder if I should describe such without vowel length, considering how questionable the status of vowel length is in GA, and with the lowering of /ɔː/ to /ɒ/, to match the actual quality it has in present-day GA, resulting in the following system:

/æ ɛ eɪ̯ ɪ i ə** ɑ ʌ** ɒ oʊ̯ ʊ u aɪ̯ aʊ̯ ɔɪ̯/

and the following vowels followed by /r/:

/ɛr ɪr ɜr*** ər*** ɑr ɔr ʊr/

I generally do not do so, though, to more closely match the universal vowel phoneme transcription I use when speaking of other English varieties, even if it is at the cost of representing GA not as closely as possible.

[Travis B.]

What I was really interested in here, though, is how other people choose to transcribe things, and their rationales for why they do so. In particular, I have noticed a lot of people transcribing general English phonemes, corresponding to what I refer to as "universal" above, with phonemes different from the ones I use; e.g. transcribing them without marking historical phonemic vowel length, using /e o/ instead of /eɪ̯ oʊ̯/, using /əʊ̯/ instead of /oʊ̯/, and so on.

There isn't really a standard way of doing it; the nearest to one is probably Wikipedia's. I don't really think it's very important which you use, as long as it's not too confusing (but I have to confess a dislike of those non-syllabicity diacritics: are they really necessary in the context?). Personally I prefer to use Wells's lexical sets rather than IPA symbols in most cases.

Those non-syllabicity diacritics probably are not necessary in phonemic transcription; I just ended up using them largely because in phonetic transcription my own dialect has both onglides and offglides, and I figured if I am doing it in phonetic transcription it would not hurt in phonemic transcription (as it does not exactly make the phonemic transcription unnecessarily narrow, but rather just clarifies which component is the glide).

At times I have figured that it would probably be a good idea to use Well's lexical sets, but I find myself always having to look them up whenever I want to use them. I also find myself wondering if I am actually confusing people not familiar with them, e.g. I am not sure if other people know that HAPPY, LETTER, and COMMA actually refer to the second vowels, and not the first, in those words.

For my own dialect I tend to use a hybrid of the Gimson and Upton schemes for RP (using Upton's choices for TRAP, DRESS and SQUARE but Gimson's for NURSE and PRICE, so /a/ /ɛ/ /ɛː/ /ɜː/ /aɪ/), with adaptations for ways in which my speech differs from RP.

That does not look like a bad system to use. Frankly, I get rather annoyed by systems for transcribing RP that mark what everyone else transcribes as /ɛ/ instead as /e/, as I am always used to reading /e/ as a mid-close monophthong (e.g. in Scottish English), and have to wrap my brain around the idea that it is actually being used as a mid-open monophthong despite there being a perfectly good IPA symbol for that.

[Travis B.]

I notice that Wikipedia uses /ɝ/ and /ɚ/ for transcribing what I would transcribe as /ɜr/ and /ər/, which is interesting. It also for North American English varieties does not mark vowel length, which is probably a good idea (but then, I know we have had people on here from North America who have reported having historical phonemic vowel length, so I am still not sure as to what its particular status is in NAE).