Semantic change

[ECOLING at aol.com]

In a message dated Patrick Ryan:

>Having worked extensively with language-family
>comparisons, my statistically unsupported "guess" is that there is amazingly
>little vocabulary loss if one allows reasonably semantically-shifted pairs
>to be counted as "retained".

That would be circular without further constraints,
because it would mean (not what Pat means,
but taken to extreme) that any meaning could change
into any other meaning, and so we are only then seeking
look-alikes, not cognates.

Rather, there needs to be a quantification of
degree of semantic change,
and of course a life-long learning of what meanings
ARE known to change into what other meanings
under what circumstances,
so we can even begin to try to measure
how far a purported related word is
from the meaning of another word we are comparing
it to.
Then, as a matter of degree, we can say that
as we allow the meanings to be more distant,
we get more *possible* cognates between any
two languages we wish to compare.
(for a fixed degree of phonetic resemblance we
might require, however measured).

All of these things are matters of degree.

A less plausible semantic relation means that two
look-alikes are less plausible as cognates.
But requiring "identity" of meanings or even
"near-identity" of meanings is an absurd
requirement also, when we are working at great
time depths.

So what to do?
Asserting simplistic extremes either for or against
is not particularly useful.

Lloyd Anderson
Ecological Linguistics