zompist bboard

More or less relevent to this thread, or at least the direction it has taken: in Finnish, /ŋ/ only exists intervocalically in inflected forms of words that have /nk/ [ŋk], eg. kenkä -> kengät (shoe -> shoes). Phonetically it's geminate or doubled [ŋː~ŋŋ], or at least occasionally [ŋg] for some people, but phonemically /ŋ/ even though it could be analysed as /ng/ or /nɣ/; the problem is that Finnish doesn't have /g/ except in loanwords and /ɣ/ has been lost. So, even though the syllable break is typically considered to be between the <n> and <g>, as in <ken.gät>, there is no /g/ phoneme.

Yet, even though /ŋ/ exists in Finnish and is a fairly common sound, some Finns at least occasionally have trouble with what is written <ng> in English and pronounce it as [ng], exactly as written. I'd say it either only happens or is especially common intervocalically, where it should be easier than word-finally if the reason why it happens was influence from Finnish. It might be somehow related to the fact that at least some dialects/speakers of English distinguish between [ŋ~ŋˑ~ŋː] and [ŋg], although it could just be that word-final [ng] takes more effort to pronounce than [ŋ]. On the other hand, some universally pronounce word-final /ŋ/ as [n] like is the default in certain dialects/sociolects of English. I've noticed that I do it sometimes, too, and something I really struggle with is pronouncing intervocalic [ŋ] without gemination or doubling; I'd pronounce singer as [siŋːə~siŋŋə] nine times out of ten even if I consciously made an effort to pronounce it as [siŋə].

the syllabification thing is moot anyway as you can always redefine the criteria as pre-stress and post-stress. i think if you do that then h/ŋ complementary distribution holds up

i heard a lot of phonologists don't really use phonemes and allophones as formal linguistic concepts anymore, but then the main other thing i've heard of is optimality theory so i dunno.

also rat-rap and rat-trap aren't minimal pairs, what?

There's still post-tonic /h/ in Irish/Scottish names like Bohannon, Houlihan, Hearlihy, vs /N/ in Kiplinger, Singer, etc. I've read on Wikipedia that in Ireland the "gh" of names like Foughey is also pronounced.
Edit: actually the H in Bohannon is probably stressed, don't know if that was original though. I've heard the name Meehan both with and without the /h/.

finlay wrote:also rat-rap and rat-trap aren't minimal pairs, what?

...They certainly are for me.

rat rap [ˈɹ̱ˁʷæʔ ˈɹ̱ˁʷæp̚]
rat trap [ˈɹ̱ˁʷæʔˌtʰɹ̱ˁʷæp̚]

finlay wrote:the syllabification thing is moot anyway as you can always redefine the criteria as pre-stress and post-stress. i think if you do that then h/ŋ complementary distribution holds up

I don't think it does, no - you can have /N/ before a stressed, zero-onset syllable ("painting-easel"), and you can arguably have /h/ as the onset of an unstressed syllable ("Flaherty" - I say 'arguably' because I can't think of any non-name examples right now; but of course the name/not-name distinction is weak, and some names can effectively become proper nouns (eg "he ate some Flahavan's for breakfast"))

also rat-rap and rat-trap aren't minimal pairs, what?

...they certainly are for me.

"Rat-rap" has /t.r/, in which the /t/ may sometimes be [?] (as it often is in codas), and is otherwise unreleased, and the /r/ is labiodental (as it always is in onsets).
"Rat-trap" has /t.tr/, in which the first /t/ is almost always [?], but the second /t/ is a released [t] (as it probably always is in onsets*), and the /r/ is alveolar (as it always is in alveolar same-syllable clusters).

Now, we could assume that I have two /t/ phonemes AND two /r/ phonemes, but that seems like a mess. We could say that the distinction is between /tr/ and /ttr/ - that /t/ is alveolar and released in codas only when following another obstruant (which I guess is true)... but this still doesn't explain the /r/ alternation (it wouldn't explain, say, "caltrop" vs "felt-wrap", unless we went around positing a lot of near-invisible geminates all over the place; come to think of it, the /t/ release/nonrelease would still be a problem for that pair).
Plus, a bunch of people have much more extensive coda glottalisation than I do, which would make such ad hoc solutions fail.

What might work, at least for me, is directly blaming the morpheme boundary. But it doesn't feel right to rely on morphological explanations directly for phonological alternations. To me, it seems that the morphology effects the allophony through some phonological process, and it seems really straightforward just to point out that this lines up pretty much perfectly with syllabification.


*I'm not sure about the syllabification of 'button' and the like; but then, my glottalisation there is variable too, so that may actually be appropriate

oh i mean they're not only distinguished by syllabification. there's two t's in one of them...

Salmoneus wrote:

finlay wrote:the syllabification thing is moot anyway as you can always redefine the criteria as pre-stress and post-stress. i think if you do that then h/ŋ complementary distribution holds up

I don't think it does, no - you can have /N/ before a stressed, zero-onset syllable ("painting-easel"), and you can arguably have /h/ as the onset of an unstressed syllable ("Flaherty" - I say 'arguably' because I can't think of any non-name examples right now; but of course the name/not-name distinction is weak, and some names can effectively become proper nouns (eg "he ate some Flahavan's for breakfast"))

"handheld" "hip-hop" ?