Reduced vowels

Josh Lalonde   Fri May 18, 2007 5:04 am GMT
Most descriptions of reduced vowels in English that I've seen seem to imply that there are only two possibilities: complete merger, as in Australian English, and complete distinction, as in RP. In my dialect, and in most other North American ones from what I can tell, the situation is more complex. I have a partial distinction between /@/ and /I/ in unstressed syllables:
roses /"ro.zIz/
Rosas /"ro.z@z/
Lennon /"lE.nIn
Lenin /lE.nIn/
impossible [Im."pQ.sI.bUo]
impossibility [Im%pQ.s@"bI.lI.4i]
Alexis [@."lEk.sIs]
The only minimal pairs I can think of in which I make the distinction involve a morpheme boundary. This is my analysis: I have only one reduced vowel, presumably /@/. It is realized as [@] morpheme-finally and before stressed syllables. In all other positions, it is realized as [I]. This puts the distinction between /@/ and /I/ in the same half-allophonic-half-phonemic zone as raised vs. unraised /aI/ in 'writer' ["r\6I.4@`] and 'rider' ["raI.4@`].
Does this seem like an internally consistent analysis? Does it seem accurate for North American English as a whole?
Travis   Fri May 18, 2007 7:32 am GMT
My dialect seems to have a partial phonemic merger of /I/ and /@/ in unstressed syllables which is not just a pseudophonemic distinction based on morpheme boundaries. I similarly have the merger of such in morpheme-final positions which results in the distinction between "roses" and "Rosa's". In most other cases I seem to have a merger (note that the merged pronunciation may vary depending on the following phoneme, and it often may involve free variation between [@], [1] and [I] or between [@] and syllabifying a following consonant). Some cases, though, tend to be rather fixed, such as merging to [I] before /n/ and /s/ (except morpheme-finally) and merging to [@] before approximants when said approximants aren't simply syllabified (in particular, when the following syllable is stressed).

However, I there seems to be exceptions to merger to [@] before /l/ such as "Advil" which have a clear [I] that does not alternate with [1] or [@] and which does not syllabify the follow /l/. However, these cases with /I/ are neologisms, which likely explains such. Of course, such could be analyzed as the other cases having shifted to /@/, and with a new non-merging /I/ being introduced in some neologisms in such cases.

Furthermore, I do not seem to have a merger before /k/ or /nk/ at all. For example, I have "panic" ["p_hE{~:nIk], "suffix" ["sVfIks], and "bobolink" ["bo:bo:M\I~Nk] but "muckluck" ["mVkL\@k] and "chipmunk" ["tS_hIpm@~nk]; it is notable that these words are monomorphemic, which means that this case cannot be explained away by morphemic structure.
Josh Lalonde   Mon May 21, 2007 12:20 am GMT
I think the situation probably is a little more complicated than my first analysis. I do produce [1] as a reduced vowel, but I think it has more to do with sentence stress and centralization than with the phonemes themselves. I also have a clear difference between 'panic' and 'chipmunk', but I'm inclined to ascribe it to secondary stress in chipmunk:
panic ["pE@.nIk]
chipmunk ["tSIp%mVNk]
I also forgot to include syllabic consonants in my analysis. They definitely are quite frequent in my accent. I'm not sure what the rules are yet.
Advil [ad.vIo]
pencil [pEn.sUo] (syllabic /l/ of course is vocalized)
button ["bV.?n=]
Travis   Mon May 21, 2007 5:55 am GMT
>>I think the situation probably is a little more complicated than my first analysis. I do produce [1] as a reduced vowel, but I think it has more to do with sentence stress and centralization than with the phonemes themselves. I also have a clear difference between 'panic' and 'chipmunk', but I'm inclined to ascribe it to secondary stress in chipmunk:
panic ["pE@.nIk]
chipmunk ["tSIp%mVNk]<<

Yeah, I forgot about secondary stress with "chipmunk", which is likely to be perceived actually as two morphemes in a compound rather than a single morpheme with a single primary stress.

>>I also forgot to include syllabic consonants in my analysis. They definitely are quite frequent in my accent. I'm not sure what the rules are yet.
Advil [ad.vIo]
pencil [pEn.sUo] (syllabic /l/ of course is vocalized)
button ["bV.?n=]<<

Before /n/ (aside from the case of /nk/), unstressed vowels and /@/ seem to be completely merged for me, with free variation between [I~n] and [n=] occurring (but with the former being more frequent in more careful speech). Before /r/, unstressed vowels (other than /I/) and /@/ always become [R=] except if the next vowel is stressed, where then there is free variation between [@R] and [R=] present; however, unstressed /Ir/ followed by a stressed vowel is realized as [IR] or [R=] in free variation. I seem to also often prefer a distinction between /@/ and /I/ before /m/ even when unstressed, even though such are still frequently merged as [m=] or [@~m] in free variation, particularly when more unstressed.
Travis   Mon May 21, 2007 5:58 am GMT
That should be "I seem to also often preserve a distinction between /@/ and /I/ before /m/ even when unstressed".
Josh Lalonde   Sat May 26, 2007 5:22 pm GMT
Actually, I'm going to completely revise my analysis using two reduced vowels, /6/ and /I/, merged in certain environments to [@] or [1]. /6/ occurs in morpheme-final situations and /I/ elsewhere, but the distinction is lost before stressed syllables. There are a few problems with this analysis that I haven't figured out yet:
suffix ["sV.fIks]
Sussex ["sV.s1ks]
Suffolk ["sV.f1k]
Maybe 'suffix' has secondary stress on the final syllable, and therefore unreduced /I/?
The name 'Alexis' also seems to have /6/, just like 'ahead'. Maybe it's analysed as 'a-' + 'lexis'?
I also think that the merged vowel may be devoiced before voiceless consonants, which would make it harder to hear the exact vowel quality. I know this happens in Quebec French, so there is some precedent for it.