Re: [tied] Finnish hevonen "horse"

----- Original Message -----

From: Knut

To: cybalist@yahoogroups.com

Sent: Wednesday, October 31, 2001 7:26 AM

Subject: Re: [tied] Finnish hevonen "horse"

> IE s/t   -  Ur s/t. (where t could be plain or aspirated)You find a lot of examples of this rule, in the core vocabolary, and they are very easy to find.

 

Maybe because /s/ and /t/ are among the most common phonemes cross-linguistically. They also occur very commonly in grammatical morphemes such as pronouns, word-forming affixes or inflections. For this reason they are also more likely than less common segmnents to produce false cognates.

 

> But for the time being it is very difficult to find out if the "s" is an original "s" or a softening of t. The reason for this is simply that the transformation t>s is very common, but often it is partial, establishing variants of the same element. Then by analogy one of  the variants often are suppressed, reestablishing the original t, or making s universal.

 

OK, then what you get is a variable correspondence. This happens, but you can't build a case for genetic relationship on it. You should find other correspondences that are _not_ so capricious. If you find that all your correspondence sets are only "partially regular" and each fully regular subset is very small, then you are probably deceiving yourself. They can't _all_ vary at the same time. Of course other ideas are worth trying. Comparison with another possibly trelated group might reveal the source of the irregularity (a phonemic contrast or structural condition lost in PIE and PU but preserved, say, in Dravidian) and make your correspondence strict. The point is anyways that a variable correspondence of the type you describe cannot be regarded as satisfactory.

> But my whole point is as follows; By looking at too exact correspondances, you easily overlook genetic relationships.

Well, by drawing premature conclusions from inexact correspondences you easily "discover" spurious genetic relationships.

 

Piotr