*a/*a: ablaut

There are a number of PIE stems which present clear evidence for *a/*a:
ablaut, and in which the presence of an internal laryngeal is unlikely
or impossible. Roots with this ablaut typically distinguish the "strong"
grade *a: and the "weak" grade *a (though their distribution may be
disturbed by analogy) but have no zero grade. Let me discuss a
characteristic example:

*wa(:)g^- 'break (hit, smash, etc.), get broken' is listed as *weh2g(^)-
in LIV, but the analysis of the vowel quality as due to laryngeal
colouring is anything but satisfactory. A root like *//weh2g^// could be
expected to produce zero-grade derivatives with *uh2g^- and some o-grade
forms like perf. *we-woh2g^-e. No such forms are attested anywhere. We
have the Gk. present (*w)ágnu:mi 'I break' and perf. éa:ge 'is broken' <
*we-wá:g^-e, Skt. vájra- < *wág^-ro- 'thunderbolt' (a substantivised
adjective). Forms like *wag^- should not exist at all, as there is no
way they could be derived from any allomorph of *//weh2g^//. Such
inconvenient forms, which LIV explains away, one by one, as "analogical
innovations", are found in several branches, while the expected reflexes
of *woh2g^- and *uh2g^- don't occur at all.

Similar difficulties beset adjectives like Lat. glaber, Slavic *gladUkU
'smooth'. If the latter reflects *gHlah2dH-u-, where does the short /a/
of <glaber> and OHG glat come from? The zero grade of *-lah2- would be
*-l.h2-, which would have given Latin -la:- and Germanic -ul-. Then we
have nouns like *wá:stu- 'dwelling, settlement' (Ved. vá:stu-, Toch.A
was.t, B ost) ~ wástu- (Gk.(w)ástu); the variants look like relicts of
an acrostatic paradigm with *a/*a: ablaut, and again a reconstruction
like *wah2stu- can't account for the existence of forms with short *a,
or the conspicuous absence of a zero-grade *uh2.

Piotr