Are there non-rhotic English speakers who distinguish between [ən] and syllabic [n]?

Question

I'm aware that [ən] can be reduced to [n̩] in some circumstances.

Does this possibility also apply to [ən] that comes from former [ɚn] in non-rhotic English dialects? I have had limited contact with non-rhotic English (mostly television), but I only recall hearing eastern as [ˈiːstən] rather than [ˈiːsʔn̩] ¹, while words like prison are commonly heard with either [ən] or [n̩].

The only information I could find was these blog comments:

army198727: Is there anyone who consistently distinguishes /ˈsev.n/ "seven" from /ˈsev.ən/ "Severn"? Or is the transcription of the former as /ˈsevn/ in most dictionaries another example of a distinction without a distinction?

mallamb27: Yes I do, and /ˈbɪtn/ 'bitten' from /ˈbɪtən/ 'bittern', /ˈpætn/ 'patten', 'paten' from /ˈpætən/ 'pattern'. You may find this hard to believe, but I am always banging on about the upper limit of distinctive realization, and this is what makes this distinction justifiable. But only for a probably quite restricted population of consistent speakers. [...]

Do we need to add a new phoneme to the list of English phonemes (as shown by the minimal pairs bitten-bittern, pattern-patten, seven-Severn)? Or is it just an occasional idiosyncratic or artificial distinction?

1. Maybe the former is used in non-TV registers? Maybe I haven't paid enough attention?

Answer 1

I think we should take the report of contrast at face value. The consequences of that fact are not obvious, though. Before leaping to a conclusion, you have to set out your analytic axioms, the most relevant issue being what definition of "phoneme" you are operating with. The most widespread understanding of "phoneme" is that a sound is a phoneme if it cannot always be derives from something else, for example [ɾ] in American English can always derive from /t/ or /d/ – whereas, [ʃ] is a phoneme because it can't always be derived from /s/ (wash, shore) even though it sometimes can (face, facial).

Assuming that view of phoneme, the primary question would be, are m̩ n̩ l̩ phonemes of English? There are decent reasons to believe that the syllabic sonorants derive from schwa plus sonorant, especially when a full vowel alternates with schwa owing to paradigmatic stress shifts. See for example calorie ~ caloric where the first syllable alternates between stressed [ˈkæl] and unstressed [kl̩]. However, one can plausible derive many instances of syllabic sonorant from an otherwise unsyllabifiable CC cluster, thus /bɒtl/ "bottle", /rɒtn/ "rotten" are possible analyses.

Since there is reportedly no difference between "Peter" and "pita", there would be no reason to posit underlying /r/ in one word and not the other. So if "bitten" comes from /bɪtən/, then what would [bɪtən] "bittern" come from? Perhaps /bɪtərn/, since even though we don't have a reason to posit /r/ in "Peter", we sort of do in "bittern" (it helps us maintain a smaller phoneme inventory). But another analysis is at hand: "bottle" is /bɔtl/, "cotton" is /kɒtn/, and in this dielect there is no reduction of schwa plus sonorant in that dialect – schwa is entirely gone. No new phoneme is added (schwa is already a phoneme); what is "added" is a kind of consonant cluster that wasn't there earlier.

IMO the answer really hangs on the strength of evidence for reduction of /ən/ to [n̩] in that dialect, in particular, the evidence that there is an underlying vowel before the nasal. I suspect that my alternation "canon" [ˈkænən] ~ "canonical" [kn̩̩ˈnɑnɪkl̩] is not found in that dialect of English and instead you have schwa in the first (unstressed) syllable.