Re: [tied] Reconstructing a future language

From: Piotr Gasiorowski
Message: 18402
Date: 2003-02-03

----- Original Message -----
From: <tgpedersen@...>
To: <>
Sent: Monday, February 03, 2003 2:02 PM
Subject: Re: [tied] Reconstructing a future language

> The algorithm works OK when applied to attested and contemporary IE
languages. It groups them as we know they're supposed to be grouped.
But when they add reconstructed PIE to the mix, as said before, PIE
ends up somewhere near the root of the group of Romance languages. So
what you're saying is that the "innovations" that PIE shares with the
Romance languages are misleading since PIE is not a Romance language?
If so, you are agreeing with the authors; it's exactly their point.
The question is: why does PIE falsely share these innovations wiith
the Romance languages? Because they share the same habitat (and I'm
only half joking)?

It's not only common innovations but also common retentions that are counted in similarity matrices (cladistic analysis absolutely rules out clusters based on common archaisms, see the links posted by Juha). Any language conservative in some crucial respects will appear to be a "favourite daughter" in an analysis that merely counts shared traits (this is why PIE appears to be closer to Latin and Italian _within_ Romance than they all are to French). In the experiment described by the McMahons and Lohr some rather conservative languages (e.g. Lithuanian, Sanskrit, Avestan) are missing (not to mention the act that Hittite, Tocharian and Armenian are not included either), so the _selection_ is seriously skewed in the first place.

The modern branches are relatively close-knit, and little doubt if this is confirmed by any analysis that isn't entirely screwed up.