Re: [tied] How to prepare **udon soup

On Thu, 22 Nov 2001 11:47:56 +0100, "Piotr Gasiorowski"
<gpiotr@...> wrote:

> > [Miguel:] If you have /a/ and /a:/, the tendency is for one to front, and the other to back. Cf. Germanic, Baltic, Slavic, Hungarian, etc. etc.
>
> ... or in many Modern English accents, for that matter. It is also common for the short vowel to remain relatively unchanged and the long one to shift away (a: > O: in Early Middle English and dialectally in Baltic; a: > E: as part of the Great Vowel Shift, or in Attic Greek in most positions).

Hungarian is interesting because it leaves the long vowel unchanged
(<á> = /a:/), while backing the short one (<a> = /O/). In standard
Dutch, the difference between <a> and <aa> is one of short back /A/
vs. (half-)long mid /a./. However, the Amsterdam dialect (and others,
the feature appears to have spread from Brabant to Utrecht to
Amsterdam) has the reverse: <a> is /a/, <aa> is /A:/.

> Miguel's pre-PIE changes look plausible to me, except in one important respect: I find a system with no high vowels typologically implausible, even as a transitional stage. Whenever /i/ and /u/ (or their long counterparts) vacate their slots in any system (as a result of lowering, diphthongisation or whatever), other vowels _immediately_ rise to fill the empty positions. It doesn't matter if you declare vocalic [i] and [u] to be allophones of /j/ and /w/. Such allophones at least (or independent high-vowel phonemes) should exist in a vowel inventory that has diphthongs like [ai] and [au]. Even a really rudimentary vertical system like Kabardian, in which there are no diphthongs and monophthongs have given up whatever colour they once had to the consonantal shell of the syllable, retains minimally a [+/- low] contrast.
>
> I must say Miguel's explanation of ablaut and stress patterns looks convincing, but he does seem to me at times to stretch typological plausibility while striving for formal elegance. The evolution of "his" vocalism begins with a triangular arrangement of vowels -- so far, so good, few things could be less common -- and culminates in the geometrically perfect "classical PIE" system; however, the intermediate inventory (*a, *a:, *ai, *au, *a:i, *a:u) is embarrassingly "marked" despite its fearful symmetry.

Thanks.

The loss of short *i and *u at some stage in pre-PIE is something that
is imposed on us by the facts (or at least, by a reasonable
extrapolation from the facts: it is reasonable to assume that if PIE
roots lack the vowels /i/ and /u/, as they (almost) do, this is not
because /i/ and /u/ were missing from the very beginning [whenever
that was], but because they were lost at some point). The loss of the
high vowels is in itself a reasonably common thing to happen: it
occurred in Slavic, in Tocharian, presumably in NW Caucasian, and it's
occurring right now in Japanese. Whether the empty slots are *always*
*immediately* filled up by other vowels, I wouldn't dare say.
Kabardian (and NW Caucasian in general) would seem to be a
counterexample, with their plethora of labialized and palatalized (and
even palato-labialized) consonants [presumably from *Ci > *C^& and *Cu

> *CW&], and their two-vowel system *& ~ *a(:) [analyzable as

underlying /a/ ~ /a:/, as Pa:n.ini would like].

Certainly, in the *majority* cases one would expect the slots to be
filled *pretty soon*. So maybe I can change the order of the
developments, by putting the loss of short *i and *u at the same time
as Nullstufe. We would have:

stressed unstressed
*í > *(y)á *i > *(y)
*ú > *(w)á *u > *(w)
*áy = *áy *ay > *i
*áw = *áw *aw > *u
*á = *á *a > *0
*á: = *á *a: > *a

I don't think (without having thought it through too much) there are
any compelling reasons to put the loss of short *i and *u _before_
Nullstufe, as in my previous presentation. One just has to be careful
to prevent the unstressed *ay's and *aw's from feeding into the
zero-gade law as *i and *u (but the same obviously goes for *a: and
*a).

=======================
Miguel Carrasquer Vidal
mcv@...