But the Slavs would not have retained the original pronunciation of (short) *a in such an early borrowing. The normal development of *spaka- would have been Slavic *spokU; vowel epenthesis in "s + stop" clusters is unheard of in Slavic. I agree that *spaka- and <sobaka> are attractively similar (which is why many linguists regard the Iranian borrowing theory as plausible), but difficulties are difficulties and must be faced rather than shrugged off. I think a non-IE intermediary is a virtual necessity here, especially in view of the fact that borrowing from _Old_ Iranian into Russian is out of the question, while later Iranian languages show reflexes of the 'dog' word that don't look like <sobaka> at all (e.g. Farsi sag).
Cross-family comparisons involving PIE *k^won- (some of which were mentioned by João) may also appear attractive but have to be rejected if the only motivation for them is similarity. BTW, actual cognates tend to be rather _dis_similar (cf. sag : hund). Some Altaicists consider Japanese inu and Turkic *it/*yt to be related, but at least this relationship is supported by some plausible-looking sound correspodences. The reconstructed Old Chinese 'dog' word is so similar to *k^won- that it's often quoted as a candidate for an old IE (Proto-Tocharian?) loan in Chinese together with a few other words, such as 'milk' and 'honey'. This is not impossible, though hard to prove. I'm sceptical of attempts to reconstruct a 'dog' etymon beyond PIE, given how variable those 'doggy' terms are even within IE, and how easily they are borrowed.
Intercontinental megalocomparison makes even less sense. Proponents of Proto-World reconstruct *kwan 'dog' (if I remember aright). Mumbo-jumbo aside, the real basis for this impressionistic "reconstruction" (mass comparison is not a reconstructive procedure, in fact) is PIE *k^won- plus a number of accidentally similar forms from arbitrarily chosen families and languages. Of course, anything that is _not_ similar is ignored. Thus, for example, Proto-Oceanic *nkaun (an up-to-date reconstruction ought to be *gaun) is offered as a cognate despite the fact (known to Austronesianists, but not to lay visitors to Proto-World websites or readers of popular articles) that the very existence of a common Oceanic term for 'dog' is somewhat problematic and that *gaun is at best a POc innovation that replaced a Proto-Austronesian word (*wasu) lost in Oceanic. PAN *wasu is not mentioned; neither are lots of other 'dog' words just because they have no dorsals or nasals in them. Since the motto "Seek, and ye shall find" works very well if only you have a sufficient number of dictionaries and word lists to hand, the results are guaranteed to be positive, and who cares if the reconstruction is strangely IE-centric or if domestic dogs are rather unlikely creatures to have existed in the Proto-World dreamtime.
Piotr
----- Original Message -----
From: "george knysh" <gknysh@...>
To: <cybalist@yahoogroups.com>
Sent: Sunday, October 06, 2002 6:24 PM
Subject: Re: [tied] Let dogs have their day too
> ******GK: Just a follow up. If we assume that "spaka"
> was borrowed directly at a very early time, and
> acclimatized itself as just plain "spaka" (or "sbaka")
> why couldn't this "spaka/sbaka" have naturally
> developed into a pronunciation of "sobaka" in an
> environment which favoured full-voicing? I realize
> that there are other "sp" words which did not turn
> into "s&p" words, so this looks like ad hoc pleading.
> But the damn words look so close...*****