Zompist Bboard Again

Zju wrote: ↑Fri Dec 31, 2021 4:28 pm What is this syllabic /r/ you're all discussing? Doesn't <Vr> merely represent /ɚ ɝ/ in rhotic vernaculars? [kl̩t], too, doesn't sound like anything I've heard online.

Syllabic /r/ for me is always [ʁ̩ˤ], except when following a rounded vowel or /w/, where then it is [ʁ̩ˤʷ]. It does not vary in realization depending on stress.

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm

zompist wrote: ↑Fri Dec 31, 2021 8:30 am

bradrn wrote: ↑Fri Dec 31, 2021 8:12 am

I see no syllabic consonant in /ˈkʌlt/, and no stressed syllabic consonant in /ˈvʌltʃr̩/. You’re quite right about /ˈbr̩d/, /dəˈmr̩d/, /ˈnr̩vəs/ though. I’m not quite sure about adultery, since it could be either /əˈdʌltr̩ˌj̩/ or /əˈdʌltr̩ˌiː/. However, I note that there are no unambiguous examples which do not involve /r̩/.

Well, you're talking about Australian English then. For me it's [kl̩t, vl̩tʃr̩].

(I can pronounce [kʌlt] as a phonetic exercise, but there is only one sound between the k and t in cult. By contrast kilt is [kɪɫt].)

I used phonemic transcription for a reason — it comes out more as [kɔwt] for me. It’s definitely not /kl̩t/ [kut] though.

EDIT: I’ve found what seems to be an example in my dialect: the surname Milner, pronounced /ˈml̩nr̩/ [ˈmuˑnɐ]. But I’m not sure how good this example is, since for me it can also be /ˈmɪlnr̩/ [ˈmɪwnɐ].

I use phonemic transcription for /l/, as real lateral /l/, in the form of [ʟ̞] for me only really shows up initially in stressed or initial syllables. Otherwise it is always vocalized (but usually remains unmerged with a preceding vowel except in /ɔːl/, which in some words may be reduced to [ɒ]~[a]~[o̞] rather than being realized as the expected [ɒo̯]).

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm

I mean the HAPPY-vowel.

It's just [i] for me. Do you have a different vowel in happiness?

Well, it’s just [i] for me also. But it’s a strange vowel — for a start, it’s [ɪ] for a number of people, making it basically the only short vowel occurring in unambiguously open syllables. See also Wikipedia:

The phonemic status of this [i] is not easy to establish. Some authors consider it to correspond phonemically with a close front vowel that is neither the vowel of KIT nor that of FLEECE; it occurs chiefly in contexts where the contrast between these vowels is neutralized, implying that it represents an archiphoneme, which may be written /i/. Many speakers, however, do have a contrast in pairs of words like studied and studded or taxis and taxes; the contrast may be [i] vs. [ɪ], [ɪ] vs. [ə] or [i] vs. [ə], hence some authors consider that the happY-vowel should be identified phonemically either with the vowel of KIT or that of FLEECE, depending on speaker.

I identify it with FLEECE, because my variety is clearly descended from a variety that had completely lost phonemic vowel length, and it is always realized identically with stressed FLEECE except when it is reduced to [j] before another vowel.

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm

Travis B. wrote: ↑Fri Dec 31, 2021 2:42 pm For me (an NAE-speaker) like Zompist I can have syllabic /r/ in stressed syllables which is indistinguishable from unstressed syllabic /r/.

For me they are different — bird is [bəːd] whereas scabbard is [ˈskæbəd]. I begin to suspect that these are two different phonemes, merged in NAE.

The analysis of stressed syllabic /r/ is odd for me, because when before a vowel is has both a syllabic /r/ and an intervocalic non-syllabic /r/, implying that there is a separate vowel phoneme from the /r/ present. (Syllabic /l/ behaves similarly). Also, in unstressed initial syllables syllabic /r/ can separate into a non-syllabic /r/ and a schwa, implying that phonemically a schwa was already present. (Furthermore, in the dialect here, the word forget/forgot/forgotten has the schwa follow the non-syllabic /r/.) Yet at the same time, if a schwa were present phonemically in stressed syllabic /r/, it would be the only case in which a schwa could occur in a stressed syllable. This vowel cannot be chalked up to be /ʌ/ because there is a contrast IMD between curt [ˈkʰʁ̩ˤʔ] and cart [ˈkʰʌʁˤʔ] and (in informal speech) between burr [ˈb̥ʁ̩ˤ(ː)] and butter [ˈb̥ʌːʁˤ].

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm

Zju wrote: ↑Fri Dec 31, 2021 4:28 pm What is this syllabic /r/ you're all discussing? Doesn't <Vr> merely represent /ɚ ɝ/ in rhotic vernaculars?

Yes, exactly — [ɚ] and [ɹ̩] are basically the same thing. I write /r̩/ because it’s easy to type, and to emphasise that it’s phonemic representation.

(Tangentially related: I’ve come to believe that IPA is unsuited for phonemic transcription, and that we should all be using Americanist notation for phonemic transcription and IPA for phonetic transcription. This seems to be an unpopular opinion.)

I do not use /ɚ ɝ/ for describing my own speech because these both behave like phoneme pairs for me (as explained above) and also because I don't consider my own speech as having true "rhotic vowels" (it has r-colored vowels, but the vowels themselves have no actual rhoticity to them).

Travis B. wrote: ↑Fri Dec 31, 2021 6:37 pm The analysis of stressed syllabic /r/ is odd for me, because when before a vowel is has both a syllabic /r/ and an intervocalic non-syllabic /r/, implying that there is a separate vowel phoneme from the /r/ present. (Syllabic /l/ behaves similarly). Also, in unstressed initial syllables syllabic /r/ can separate into a non-syllabic /r/ and a schwa, implying that phonemically a schwa was already present. (Furthermore, in the dialect here, the word forget/forgot/forgotten has the schwa follow the non-syllabic /r/.) Yet at the same time, if a schwa were present phonemically in stressed syllabic /r/, it would be the only case in which a schwa could occur in a stressed syllable. This vowel cannot be chalked up to be /ʌ/ because there is a contrast IMD between curt [ˈkʰʁ̩ˤʔ] and cart [ˈkʰʌʁˤʔ] and (in informal speech) between burr [ˈb̥ʁ̩ˤ(ː)] and butter [ˈb̥ʌːʁˤ]. … I do not use /ɚ ɝ/ for describing my own speech because these both behave like phoneme pairs for me (as explained above) and also because I don't consider my own speech as having true "rhotic vowels" (it has r-colored vowels, but the vowels themselves have no actual rhoticity to them).

I fail to see your reasoning for saying that ‘[t]he analysis of stressed syllabic /r/ is odd for me’. How does the epenthesis of a consonant between stressed /r̩ l̩/ and a following vowel change anything here? It seems to me that [ˈkʰʁ̩ˤʔ] and [ˈb̥ʁ̩ˤ(ː)] can be analysed straightforwardly as /kr̩t/ and /br̩/. But your comment that ‘in unstressed initial syllables syllabic /r/ can separate into a non-syllabic /r/ and a schwa’ is interesting. Could it be that this ‘syllabic consonant’ is really /ər/? But that still doesn’t explain the sesquisyllabic pattern or the presence of stressed /ə/…

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm I used phonemic transcription for a reason — it comes out more as [kɔwt] for me. It’s definitely not /kl̩t/ [kut] though.

For me they are different — bird is [bəːd] whereas scabbard is [ˈskæbəd].

I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

I'm not sure that it even makes sense to try to come up with a phonemic analysis that fits all of English, as opposed to a particular dialect. Or to be precise, it can be useful for historical analysis, but history is not available to individual language learners; and as a wider issue, isn't available for discussing the vast majority of languages.

(Tangentially related: I’ve come to believe that IPA is unsuited for phonemic transcription, and that we should all be using Americanist notation for phonemic transcription and IPA for phonetic transcription. This seems to be an unpopular opinion.)

I don't think that's a bad idea. I'd just add that in many areas there is a different notation, not Americanist, used for phonemic transcriptions, especially when there's a script involved and it's desired to keep close to it. (E.g. for Biblical Hebrew almost everyone uses a close transcription of the Masoretic Hebrew, and in Indic studies the IAST is used for Sanskrit.)

bradrn wrote: ↑Fri Dec 31, 2021 6:18 am In fact, I can’t think of any good arguments against this analysis. Can anyone think of a convincing reason why English native vocabulary is not sesquisyllabic?

Firstly, I don't like it because I'm used to the iambic minor-major pattern for sesquisyllables, whereas English is closer to the Proto-Uralic pattern of major-minor - trochees. Secondly, just as these trochees seem to be allowed when required to talk in words of one syllable, are not words whose first syllable is (unstressed) /ə/ also allowed?

bradrn wrote: ↑Sat Jan 01, 2022 7:28 am

Travis B. wrote: ↑Fri Dec 31, 2021 6:37 pm The analysis of stressed syllabic /r/ is odd for me, because when before a vowel is has both a syllabic /r/ and an intervocalic non-syllabic /r/, implying that there is a separate vowel phoneme from the /r/ present. (Syllabic /l/ behaves similarly). Also, in unstressed initial syllables syllabic /r/ can separate into a non-syllabic /r/ and a schwa, implying that phonemically a schwa was already present. (Furthermore, in the dialect here, the word forget/forgot/forgotten has the schwa follow the non-syllabic /r/.) Yet at the same time, if a schwa were present phonemically in stressed syllabic /r/, it would be the only case in which a schwa could occur in a stressed syllable. This vowel cannot be chalked up to be /ʌ/ because there is a contrast IMD between curt [ˈkʰʁ̩ˤʔ] and cart [ˈkʰʌʁˤʔ] and (in informal speech) between burr [ˈb̥ʁ̩ˤ(ː)] and butter [ˈb̥ʌːʁˤ]. … I do not use /ɚ ɝ/ for describing my own speech because these both behave like phoneme pairs for me (as explained above) and also because I don't consider my own speech as having true "rhotic vowels" (it has r-colored vowels, but the vowels themselves have no actual rhoticity to them).

I fail to see your reasoning for saying that ‘[t]he analysis of stressed syllabic /r/ is odd for me’. How does the epenthesis of a consonant between stressed /r̩ l̩/ and a following vowel change anything here? It seems to me that [ˈkʰʁ̩ˤʔ] and [ˈb̥ʁ̩ˤ(ː)] can be analysed straightforwardly as /kr̩t/ and /br̩/. But your comment that ‘in unstressed initial syllables syllabic /r/ can separate into a non-syllabic /r/ and a schwa’ is interesting. Could it be that this ‘syllabic consonant’ is really /ər/? But that still doesn’t explain the sesquisyllabic pattern or the presence of stressed /ə/…

Another thing is that I am hesitant to declare a phonemic distinction between two different /r/'s based purely on syllabicity. For instance curry [ˈkʰʁ̩ˤːʁˤi(ː)] and Cree [ˈkʰʁˤi(ː)] would become simply /ˈkr̩i/ versus /ˈkri/. Of course, a way around it is to declare the former /ˈkr̩ri/, but that feels like an ad-hoc solution, no better than a restricted-distribution stressed /ə/. Of course one could not have phonemic syllabicity at all and declare it /ˈkrri/, but that doesn't explain why in some environments [ʁ̩ˤ] patterns with /ər/ by not infrequently becoming [əʁˤ] or vice versa, implying an analysis as /ər/ as mentioned or even some other vowel getting reduced (e.g. in orangutan [ʁ̩ˤːˈʁˤẽ̞ːŋɘˌtʰẽ̞(ː)ŋ]~[əːˈʁˤẽ̞ːŋɘˌtʰẽ̞(ː)ŋ]~[ɔːˈʁˤẽ̞ːŋɘˌtʰẽ̞(ː)ŋ]). In the end I would tend towards the analysis as /ər/ as the simplest for my lect, even though it requires a special stressed /ə/ before /r/.

zompist wrote: ↑Sat Jan 01, 2022 9:20 am I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm

zompist wrote: ↑Sat Jan 01, 2022 9:20 am I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

Travis B. wrote: ↑Sat Jan 01, 2022 3:04 pm

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm

zompist wrote: ↑Sat Jan 01, 2022 9:20 am I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

And intrusive /j/. Probably also intrusive /m/ and /n/ as well.

Richard W wrote: ↑Sat Jan 01, 2022 3:14 pm

Travis B. wrote: ↑Sat Jan 01, 2022 3:04 pm

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm
Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

And intrusive /j/. Probably also intrusive /m/ and /n/ as well.

Here syllabicity of unstressed final [m] and [n] is entirely optional (except of [n] in /Vtn/, which is mandatory) and extremely frequently not present, implying that these are underlying /əm/ and /ən/, which in turn implies an analysis of /ər/ and /əl/ as well. Of course that leaves unstressed /i/ as the odd one out, but that does not eject a semivowel before a vowel here but rather readily reduces to [j] or even affrication or palatalization of a preceding consonant in that case, implying that it is different from the others in the first place.

Affrication is something I would not expect in nuttier or easier or eddying. This is unlike the treatment of Latinate words or the abutment of words.

Richard W wrote: ↑Sat Jan 01, 2022 4:34 pm Affrication is something I would not expect in nuttier or easier or eddying. This is unlike the treatment of Latinate words or the abutment of words.

Tis true. Those three words don't affricate for me, and in general /t/ or /d/ realized as flaps never seem to undergo affrication in NAE. (However, unstressed /di/ between vowels IMD has a tendency to become [jː].)

Travis B. wrote: ↑Sat Jan 01, 2022 3:04 pm

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm

zompist wrote: ↑Sat Jan 01, 2022 9:20 am I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

That's not going to affect bird. For words in -r, it depends on whether Brad inserts an r only where there was historically one ("they were angry"), or before any word beginning with a vowel ("I saw a bug" > "I sore a bug").

zompist wrote: ↑Sat Jan 01, 2022 4:55 pm

Travis B. wrote: ↑Sat Jan 01, 2022 3:04 pm

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm
Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

That's not going to affect bird. For words in -r, it depends on whether Brad inserts an r only where there was historically one ("they were angry"), or before any word beginning with a vowel ("I saw a bug" > "I sore a bug").

IIRC Australian English has intrusive /r/; IIRC the only non-rhotic English varieties that lack intrusive /r/ are non-rhotic NAE varieties and South African English.

Travis B. wrote: ↑Sat Jan 01, 2022 5:28 pm IIRC Australian English has intrusive /r/; IIRC the only non-rhotic English varieties that lack intrusive /r/ are non-rhotic NAE varieties and South African English.

We're getting away from native 'sesquisyllabic' words. There are certainly British near-RP lects that lack word-internal intrusive /r/, e.g. in words such as _drawing_.

Richard W wrote: ↑Sat Jan 01, 2022 6:53 pm

Travis B. wrote: ↑Sat Jan 01, 2022 5:28 pm IIRC Australian English has intrusive /r/; IIRC the only non-rhotic English varieties that lack intrusive /r/ are non-rhotic NAE varieties and South African English.

We're getting away from native 'sesquisyllabic' words. There are certainly British near-RP lects that lack word-internal intrusive /r/, e.g. in words such as _drawing_.

My point was that native "sesquisyllabic" words in English, Australian, and NZ non-rhotic varieties cannot be really considered as having a final /r̩/ but rather as having a final /ə/ that just happens to undergo intrusive /r/ when followed by a vowel-initial word. Positing a final /r̩/ separate from final /ə/ in a non-rhotic variety only makes sense if they have a linking /r/ without an intrusive /r/, something which I have heard is rare to nonexistent.

I was referring to the behaviour when suffixes beginning with a vowel are added to a word ending in a syllabic resonant.

Speakers of non-rhotic English with lexically specified linking-r aren't all dead yet. By Bradn's implicit description, the words ending in /ə/ without linking-r aren't native. I'm not sure what the status of cuppa is - perhaps he will deny that it can lack linking-r in what he regards as English. @Bradn, what is your analysis of yellow?

zompist wrote: ↑Sat Jan 01, 2022 9:20 am

bradrn wrote: ↑Fri Dec 31, 2021 5:20 pm I used phonemic transcription for a reason — it comes out more as [kɔwt] for me. It’s definitely not /kl̩t/ [kut] though.

For me they are different — bird is [bəːd] whereas scabbard is [ˈskæbəd].

I hadn't realized your /l̩ r̩/ were quite so vocalic. For your dialect, I'm not sure what the sesquisyllabic idea adds to the analysis? It sounds like you just have ordinary vowels (or diphthongs) for these, not consonantal vowels.

I came up with the sesquisyllabic analysis when I realised that a disproportionate number of disyllabic native words end with one of [m̩ n̩ u ɐ i] — surprisingly few end with anything else. [u] is transparently /l/, and [ɐ] corresponds to [ɹ̩] in rhotic dialects, and from there the sesquisyllabic analysis as /m̩ n̩ l̩ ɹ̩ j̩/ becomes obvious.

I'm not sure that it even makes sense to try to come up with a phonemic analysis that fits all of English, as opposed to a particular dialect. Or to be precise, it can be useful for historical analysis, but history is not available to individual language learners; and as a wider issue, isn't available for discussing the vast majority of languages.

This is probably true. On the other hand, we are doing phonemic analysis, so the historical information can be valuable here. And the sesquisyllabic phenomenon isn’t restricted to any one dialect of English.

Richard W wrote: ↑Sat Jan 01, 2022 2:26 pm Firstly, I don't like it because I'm used to the iambic minor-major pattern for sesquisyllables, whereas English is closer to the Proto-Uralic pattern of major-minor - trochees.

Either way, there’s still one major and one minor syllable in the word, and the minor syllables seem to have similar properties across both patterns, which for me is enough to justify a sesquisyllabic analysis.

Secondly, just as these trochees seem to be allowed when required to talk in words of one syllable, are not words whose first syllable is (unstressed) /ə/ also allowed?

I don’t understand this comment. Could you elaborate please?

Travis B. wrote: ↑Sat Jan 01, 2022 2:31 pm Another thing is that I am hesitant to declare a phonemic distinction between two different /r/'s based purely on syllabicity. For instance curry [ˈkʰʁ̩ˤːʁˤi(ː)] and Cree [ˈkʰʁˤi(ː)] would become simply /ˈkr̩i/ versus /ˈkri/ … In the end I would tend towards the analysis as /ər/ as the simplest for my lect, even though it requires a special stressed /ə/ before /r/.

For me, curry is [ˈkʰɐɻʷi]. This seems a genuine dialectal difference: my /kʌrj̩/ vs your /kərj̩/.

Richard W wrote: ↑Sat Jan 01, 2022 4:34 pm Affrication is something I would not expect in nuttier or easier or eddying. This is unlike the treatment of Latinate words or the abutment of words.

I have [ˈnɐtˢijɐ], but that’s just because my /t/ is regularly [tˢ].

zompist wrote: ↑Sat Jan 01, 2022 4:55 pm

Travis B. wrote: ↑Sat Jan 01, 2022 3:04 pm

Richard W wrote: ↑Sat Jan 01, 2022 2:38 pm
Morphophonemics. When a suffix beginning with a vowel is added, final /r̩/ can split to /r̩r/.

An alternate analysis, for bradrn's lect, would be that this is simple intrusive /r/, and by analogy, intrusive /l/.

That's not going to affect bird. For words in -r, it depends on whether Brad inserts an r only where there was historically one ("they were angry"), or before any word beginning with a vowel ("I saw a bug" > "I sore a bug").

To some extent there’s free variation, but I can insert it everywhere: [ɑ͡iˌso̞(ɻʷ~w)əˈbaɡ]. In this case the variant with intrusive [w] seems slightly more natural, I think. But as I said, morphophonemics was not my main motivation for this analysis.

Richard W wrote: ↑Sat Jan 01, 2022 8:07 pm Speakers of non-rhotic English with lexically specified linking-r aren't all dead yet. By Bradn's implicit description, the words ending in /ə/ without linking-r aren't native. I'm not sure what the status of cuppa is - perhaps he will deny that it can lack linking-r in what he regards as English.

I’m not quite sure either — it doesn’t tend to precede vowel-initial words: a cuppa, cuppa tea, cuppa coffee. But the non-idiomatic combination cuppa alcohol I would pronounce with a linking-r: [ˌkʰɐpʰəˈɻʷæwkʰəˌhow].

@Bradn, what is your analysis of yellow?

I analyse /ˈjelə͡u/ as a truly disyllabic word, not sesquisyllabic at all.

bradrn wrote: ↑Sat Jan 01, 2022 9:00 pm

Richard W wrote: ↑Sat Jan 01, 2022 8:07 pm Speakers of non-rhotic English with lexically specified linking-r aren't all dead yet. By Bradn's implicit description, the words ending in /ə/ without linking-r aren't native. I'm not sure what the status of cuppa is - perhaps he will deny that it can lack linking-r in what he regards as English.

I’m not quite sure either — it doesn’t tend to precede vowel-initial words: a cuppa, cuppa tea, cuppa coffee. But the non-idiomatic combination cuppa alcohol I would pronounce with a linking-r: [ˌkʰɐpʰəˈɻʷæwkʰəˌhow].

I am amused by the analysis of cuppa as a single word, since at least in NAE /ə/ is an allomorph of of across the board, and not just after cup.

Travis B. wrote: ↑Sat Jan 01, 2022 11:34 pm

bradrn wrote: ↑Sat Jan 01, 2022 9:00 pm

Richard W wrote: ↑Sat Jan 01, 2022 8:07 pm Speakers of non-rhotic English with lexically specified linking-r aren't all dead yet. By Bradn's implicit description, the words ending in /ə/ without linking-r aren't native. I'm not sure what the status of cuppa is - perhaps he will deny that it can lack linking-r in what he regards as English.

I’m not quite sure either — it doesn’t tend to precede vowel-initial words: a cuppa, cuppa tea, cuppa coffee. But the non-idiomatic combination cuppa alcohol I would pronounce with a linking-r: [ˌkʰɐpʰəˈɻʷæwkʰəˌhow].

I am amused by the analysis of cuppa as a single word, since at least in NAE /ə/ is an allomorph of of across the board, and not just after cup.

This is the case for me too. But there is evidence that this particular combination has lexicalised — a cuppa (without following noun) is valid for me, while *a cup of is not.