Re: [tied] "Fish" in Slavic

On 2006-07-22 19:38, Grzegorz Jagodzinski wrote:

> - Russian rys', Serbian/Croat ri"s, Slovene ri^s show no traces of *n,
> - Lith. lú:s^is, Latv. lu~sis, OPruss. luysis (uncertain) show no traces of
> *n, even if there are dialects that have it,

_No_ traces of *n _even_ if there are dialects that show it? This is
self-contradictory. It's precisely those dialectal (Samogitian) forms
that are relevant, so you can't simply sweep them under the carpet!

> - OHG luhs (now Luchs), Old Swedish ló < *luxsu-, *luxa- (x like in IPA)
> show no traces of *n,

In all lineages of Germanic *n was lost very early before a velar
fricative, leaving a lengthened vowel. It's true that the attested
reflexes of the 'lynx' word point to a short vowel and therefore no
nasal, but note that in Balto-Slavic, au contraire, the vowel is _long_
and there's some evidence of original nasalisation.

> - MIrish lug also is not any evidence for *n,

The Celtic form is so aberrant that it's either unrelated or
contaminated with some other etyma. In effect, it hardly counts as
useful evidence for anything.

> - Armenian lusanunk' has two -n-'s but none on the right place; in fact lus-
> < *luk^-, not *lunk^-

Still, it's a nasal stem, and metathesis in oblique allomorphs of nasal
stems occasionally produces what looks like an infix. Cf. Slavic *reNdU
and Lat. ordo:, discussed here a while ago. Let's imagine, for the sake
of the argument, that the 'lynx' word was originally something like
*luk^-é:n, gen. luk^-n-ós > *lunk^-ós, hence the analogical nom.sg.
*lunk^-s. If the same root could occur with other suffixes, we can both
account for the n-less forms with short *u and make sense of the
combined testimony of Greek, Balto-Slavic and Armenian.

Come to think of it, as this tentative reconstruction looks like an
agent noun, Skt. rus'áti 'hurts, injures' may well be related. I've seen
less plausible etymologies of 'lynx'.

> - and finally only Greek has lyn,x, gen. lyn,kos and lyn,gos (-n,- spelt
> with gamma of course)
>
> Even leaving the variation k/g in Greek (secondary and
> unetymological), why should we believe that -n- in Greek and
> (partially) Baltic is not secondary?

What could motivate such a secondary *n _ex nihilo_ independently in two
branches?

> Which is more, one could expect
> Grek **lax < **lwak- < IE *lwn.k-, rather than lynx.

For reasons given earlier, I would't expect PIE *[lwn.k] in the first
place, as the optimal syllabification was *[lunk] (after a sonorant). If
the -nk- is of metathetic origin, -unk- is all the more expected, as in
nasal presents.

> Or even **la:k-, taking into consideration the view that Baltic and
> Slavic forms must have come from *lunHk^- and not *lunk^-

That would have given **lunVk, not *la:k, in Greek. I think the length
in PBSl. *lu:nc'i- reflects a special treatment of inherited *iR/*uR
(contrasting with BSl. reflexes of syllabic liquids and nasals).

Piotr