First, Bengali has a rule that, put in a simplified manner, dictates that vowels raise before a high vowel. This is most evident in verbal systems --
do + 1P.Pres
as opposed to
do + 3P.Pres
This variation is fairly common and happens in a lot of contexts, in all syntactic categories. So looking at this set, it seems [o] is an allophone of /ɔ/. But there is a verbal form that is also a suffix pronounced "-e," and unlike the 3rd person present form of the verb, it raises the verb before it -- being historically "ije."
do + 'past active participle'
"Doing", "In doing", "by", "having done"
Both of these -e morphemes apply to all verbs and, in every case, verbs with the -e morpheme in example (2) all raise -- including /a/, which likes to disobey rules that apply to all other vowels -- whereas verbs that go through the -e morpheme in example (1) do not.
So, would ɔ and o be two different phonemes, as implies the minimal pair [kore] vs. [kɔre], or would [ɔ] and [o] be allophones of the same morpheme? I suppose it isn't completely ridiculous for the suffix in (2) to be -e with a raising feature, but it's strange in that the raising happens in an environment where it normally wouldn't.
Does anything else like this happen in other languages? If so, is there a term for it? Speakers of vowel harmony languages, does this seem completely bizarre or are there cases in which vowel harmony 'switches off' to a change in meaning?
X-posted to my journal and linguaphiles