Pre-Slavic *o would have been lengthened by a following plain *g, giving Slavic *a (at least it _probably_ would; the precise conditioning of Winter's Law is still something of a mystery.
We can get round this difficulty if we assume *n.gni- > *ungni- > *u:ngni- > *o~gnI > *ognI, with dissimilatory denasalisation, as has been proposed for <ignis> (as a bonus, we get an easy explanation of Slavic by-forms with vyg- < *u:g-; they represent a variant in which denasalisation occurred a little earlier; to Pokorny they were "completely unclear").
Baltic *u as in Lith. ugnis and Latv. uguns could then be derived in a similar from the same Proto-Balto-Slavic *ungnis (I can't take seriously the idea that they owe their /u/ to the influence of Old Swedish ugn 'oven').
To sum up, although the first *n does not surface anywhere directly, it's a hypothesis that allows us to unify the etymon.
I wouldn't be so sure that the expected initial laryngeal was *h1. *Hn.gnís is what comparative analysis plus phonotactic considerations suggest. Perhaps it's somehow related to Pokorny's 'coal' etymon (PBSl. *angli-, with Indo-Iranian and possibly Celtic cognates < PIE *h2ango:l/*h2angl- 'coal, glowing ember').
Piotr
----- Original Message -----
From: <richard.wordingham@...>
To: <cybalist@yahoogroups.com>
Sent: Wednesday, February 26, 2003 10:49 AM
Subject: [tied] *h1ngni-, Fire
> Why does one reconstruct *h1ngni- from Latin ignis, Sanskrit agni- and Proto-Slavic *ognI? I can't see the justification for the first /n/, though it is consistent with the Latin and Sanskrit forms. (So is the reconstruction of the first vowel as /e/, as in
Pokorny.)