Odd reconstructed "Vocabulary Density"
ECOLING at aol.com
ECOLING at aol.com
Thu Aug 12 03:03:13 UTC 1999
Pete Gray wrote as quoted at the end of this message.
I think the conclusion is that reconstruction (or the comparative method
or whatever one wishes to call it) should try to do a much better job
of attributing small differences in sense to reconstructed items.
It is NOT the case that the meaning "grow" is somehow more neutral
or less specific than the other meanings which are nearby in its semantic
field. Rather, our current methods, by convention, artificially push
the reconstructed meanings towards the crude set of pigeon-holes we
Just as in sociolinguistics, where it is known that meta-talk ABOUT
our own speech is much less accurate than the observable behavior
we exhibit when actually speaking. We need to have self-conscious
correctives against this recurring error endemic to the conscious analysis
The comparative / reconstructive methods, in the guise of
claiming not to be more precise than the limited data warrant,
are actually throwing away information on details of meaning.
Following quoted from Pete Gray:
(b) I can only speak for PIE amongst the reconstructed languages. Its
reconstructed vocabulary is certainly odd. Not including obvious root
(i) There are 18 roots for glisten/glitter, and 12 for shine (total 30)
(ii) There are 8 for goat
(iii) there are 8 or 9 for grow
(iv) There are 23 for hit
(v) There are 10 for jump
(vi) There are 11 for weave/plait
(vii) There are 12 for pull
(viii) There are 11 for press
(ix) There are 24 for turn
(x) There 17 for swell
There are a large number of roots with similar semantic connotations, (over
half the semantic concepts have at least two reconstructed independent
roots). Some of these have large numbers of these "pseudo-synonyms".
Given the patchy and limited nature of what we can reconstruct, it certainly
seems that reconstructed PIE has its words clustered around some concepts at
the expense of others.
So this gives two questions:
(A) Is this pattern anything like natural languages?
(B)Is the overall average anything like the overall average in natural
More information about the Indo-european