Re: Icelandic umlauts.

> Date:         Mon, 19 Jun 2000 13:00:45 -0400
> From: John Cowan <jcowan@...>
>
> Oskar Gudlaugsson wrote:
>
> > The relationship between those stems is *not* umlaut, but
> > rather a different kind of sound change called "sound shift" (my
> > translation).
>
> "Ablaut", at least when applied to the Germanic strong verbs
Ablaut is the name of a pervasive phenomenon in PIE morphophonemics
which can affect all stems both in derivation and inflection.

The strong verbs are the main place in inflection where it came to
support a semantic load and was thus maintained in a balance with
analogical pressures. IIRC, except for verb stems there's not much
inflectional ablaut that survives to historical languages outside
Indo-Iranian.

According to some scholars, the thematic vowel in some inflectional
stems underwent Ablaut as well, but if so it has been well levelled by
analogy, and the conclusion is not very certain yet.

In derivation, the Ablaut grades of the same stem often developed very
differently, in effect making different roots from a synchronic point
of view.

In the usual reconstruction of PIE, Ablaut shows up as a variation
between e, o and 0 in some syllable of a morpheme, according to its
environment (including supramorphemics like accent). Later phonetic
developments have obscured that, giving variations like in the
Germanic strong verbs.

However, the variation in PIE was likely derived from something at an
earlier stage --- one theory is that /a/ split into /a/, /a:/ and /0/
due to stress; /a:/ then > /A/ (as for instance in Arabic); and lastly
/a/ > *e and /A/ > *o. (With wide scope for any of these processes to
be blocked in certain contexts, making Ablaut somewhat irregular at
the PIE level).

(Note: This theory does not imply that there was ever a predecessor
stage where /a/ was the only vowel. In PIE as often reconstructed, *j
and *w are classed as sonorants along with *n, *m, *r, and *l (and the
laryngeals), and are grudgingly granted status as syllabic nuclei when
there isn't a 'real' vowel (*e or *o) around to make them into glides.
The idea seems to be that there must underlyingly be an *e in every
syllable, so what looks like **u is just the zero stage of *we or *ew.

In fact, that may well be a reasonable 'synchronic' description of the
reconstructed PIE. But it doesn't imply that there couldn't have been
any /i/s and /u/s at the hypothetical stage before /a/ split up. For
instance, they might have gone to zero where /a/ did, turned into
glides next to /a/, merged with /e/ or /o/ in other positions --- and
a little later, analogy might actually have turned /u/ into *we in
some positions where Ablaut was expected. All this obscures the matter
to the point where internal reconstruction can't determine whether /i/
and /u/ in fact existed as phonemes, and people are reduced to arguing
from Nostratic or typology).

Lars Mathiesen (U of Copenhagen CS Dep) <thorinn@...> (Humour NOT marked)