'Dog' revisited

The analysis of *k^wo:n as *pk^wo:n = *pk^u-o:n (derived from *pek^u-
'livestock') has already been discussed on Cybalist. It has occurred to me
that the etymology fits in very well with the known patterns of IE compound
formation. One only needs to assume that the word is really a compound in
which the second element is the "Hoffmann suffix" *-hVn-. (I deliberately
omit the laryngeal index in order to avoid getting sidetracked into
discussing a minor issue.) Transparent formations with *-hVn- can to all
intents and purposes be treated as compounds. If the first element is a stem
with ablaut allomorphs, it exhibits the nil-grade (characteristically if
oddly stressed in the "strong" cases, as in familiar compounds with *h1s�-,
*kWtwr.'- ~ *kWtr�-, *sm.'-, *n.'-, etc.). The second element is declined as
follows:

nom.sg *'-ho:n
acc.sg. *'-hon-m.
gen.sg. *-h�n-s --> *-h(e)n-o/es
loc.sg. *-h�n-i
gen.pl. *-hn-�m

A clear example of such a compound is *h2j�-hon-/*h2ju-h�n-/*h2ju-hn-�-
(from *h2oju- 'vital energy, life, youth'), becoming *j�won-/*juw�n-/*ju:n-
after the loss of the laryngeals. cf. Skt. y�van-/yu:n- 'young man'. A less
clear case is *dHg^Hm.'-hon- 'human being' (from *dH�g^Ho:m/*dHg^H�:m
'earth', with prevocalic "syllabic" *m. realised as *[m.m]), which became
*g^Hm.'-on- (OLat. hemo: > Lat. homo:, OE guma) or *g^Hm�n- (Lith.
z^muo~)Note the simplification of the initial cluster in both. The latter
variant resulted from the reinterpretation of *g^Hm.'o:n as the "Lindeman
treatment" of an underlyingly monosyllabic word (like *dij�:us as a
disyllabic realisation of *dje:us). Each stem variant spawned its own
constellation of case forms, obscuring the original pattern.

Hypothetical *pk^�-hon- could be expected to produce the following forms,
among others:

nom.sg. *pk^�-ho:n > *k^�wo:n (cf. Gk. k�o:n)
acc.sg. *pk^�-hon-m. > *k^�won-m.
gen.sg. *pk^u-h�n-s > *k^u�n(o)s etc.
gen.pl. *pk^u-hn-�m > *k^u:n�m (Cf. Av. su:nam)

Since the compound was obscured very early and the simplification of the
cluster *pk^- took place already in PIE, we have no trace of the initial *p-
even in Iranian (which normally has *fs^u- as the weak form of *pasu- <
*pek^u-). A similar case is *k^m.t�m < *dk^m.t�m, exhibiting no "thorny"
reflexes of the initial anywhere. At first blush, the short-vowelled weak
cases such as gen.sg. *k^un�s (Gk. kun�s, Lith s^une`s) or *k^�nos (Skt.
s'�nah.) might seem to rule out *-hn-. However, they difficulty disappears
when one realises that they are late secondary forms based on the new
underlying stem *{k^won-} extracted from the nom.sg. *k^uwo:n interpreted as
a Lindeman form (hence post-PIE *k^(u)w�:n, *k^w�nm., *k^un�s, *k^un�m).

We have seen that the evolution of *dHg^Hm.-hon- went along similar lines.
Why wasn't *h2j�-hon- refashioned in the same way? The reason is plain: the
phonotactics of most IE languages rule out a Lindeman reinterpretation of
*j�wo:n by prohibiting monosyllabic *jwon-. Consequently, the stem continued
to be analysed in post-PIE terms as *{juhVn} or *{juwVn-} (realised as
*j�wo:n, *j�wonm., *juw�n(o)s, *ju:n�m).

The compound analysis of the 'dog' word explains in a simple and natural way
its apparently aberrant -- actually archaic -- forms and accounts for the
observed range of variants (*k^won- fares worse in these respects). The
semantics of 'livestocker' --> 'sheepdog' --> 'dog' is not particularly
problemetic. Any thoughts?

Piotr