Twin Aster

[Darren]

At minimum I’d expect something like /əj jə əw wə/ → [e e o o]. Come to think of it, I’m not sure there’s any two-vowel system where this doesn’t happen.

I guess there's Nuxálk with allegedly just /a/ that doesn't display notable allophony (and syllabic /j w/). Maybe Arrernte too with /ɪ~ʊ a/ where /ɪ~ʊ/ variation is fairly unconditioned? And I'm tempted to say some stage of PIE (if at one point it had just */a aː/ or */æ ɒ/ or something)? I think with the syllabic semivowels it's reasonable enough.

I usually take posited "syllabic semivowels" to mean that the people doing the analysis are being too clever.

This is usually true, although I can see a justification for it in languages like PIE where ablaut plays a major role.

[Travis B.]

I guess there's Nuxálk with allegedly just /a/ that doesn't display notable allophony (and syllabic /j w/). Maybe Arrernte too with /ɪ~ʊ a/ where /ɪ~ʊ/ variation is fairly unconditioned? And I'm tempted to say some stage of PIE (if at one point it had just */a aː/ or */æ ɒ/ or something)? I think with the syllabic semivowels it's reasonable enough.

I usually take posited "syllabic semivowels" to mean that the people doing the analysis are being too clever.

This is usually true, although I can see a justification for it in languages like PIE where ablaut plays a major role.

Thing is, a vowel system of, say, */i æ ɒ u/ or */i a aː u/ is much more sane than one of just */æ ɒ/ or */a aː/ (after all, real extremely small vowel inventories are typically vertical, e.g. /a ɨ/). I feel that sometimes the people doing analyses forget that what they are analyzing were real languages spoken by real people at one point, and not just figments of linguists' imaginations.

[Darren]

Thing is, a vowel system of, say, */i æ ɒ u/ or */i a aː u/ is much more sane than one of just */æ ɒ/ or */a aː/ (after all, real extremely small vowel inventories are typically vertical, e.g. /a ɨ/). I feel that sometimes the people doing analyses forget that what they are analyzing were real languages spoken by real people at one point, and not just figments of linguists' imaginations.

Yeah I've got no argument with that. But it's also true that PIE */i u/ and */j w/ alternate in the same way that */m̩ n̩/ and */m n/ do, so there isn't really a good reason to say that */i u/ are vowels and */m̩ n̩/ aren't. It's equally as sane as claiming that say Ubykh has only two vowels /ə a/ when all of [i y ɨ u e ø ə o ɛ œ a ɔ] or whatever appear with high frequency. Sure, if you were writing a layman's grammar of PIE, you'd say it has the vowels */i u e o a/, but we do gain some understanding of the language from recognising that */i u/ (and */a/ for that matter) are not like the other vowels.

(sorry for the derailing of your thread Mr Space)