Re: Mother of all IE languages

--- In cybalist@yahoogroups.com, "Brian M. Scott" <BMScott@...>
wrote:

> At 1:18:08 on Wednesday, 3 December 2003, S.Kalyanaraman wrote:
>
> > --- In cybalist@yahoogroups.com, "Brian M. Scott"
> > <BMScott@...> wrote:
>
> >>> When words are excluded from a list by design, the
> >>> selection does not become random and to such a sample, the
> >>> statistical techniques cannot be applied.
>
> >> This is nonsense. Swadesh-like lists are not intended to be
> >> random samples and are not used as if they were.
>
> > There is nothing nonsensical about statistical techniques.
>
> I quite agree. I'm a mathematician, after all, and I
> occasionally teach basic statistics. I was commenting on
> your misunderstanding, not on the statistics.
>
> > Two definitions are critical: population and sample. If a
> > sample is used to derive the characteristics of a
> > population, the sample has to be random.
>
> Do you know the expression 'teaching your grandmother to
> suck eggs'?
>
> Whatever its shortcomings, the Gray/Atkinson technique is
> not an example of statistical inference, so your comment is
> inapplicable to it. It isn't even applicable to
> glottochronology, which -- however useless -- didn't
> pretend that a Swadesh list was a random sample or try to
> derive characteristics of the lexicon from it.

While on the whole I agree with Brian, there are some problems with
the Swadesh list. They became apparent when I was drawing up some
lists. Some pairs of words on it are often cognate - 'thou'
and 'you' (the same word in quite a few languages, e.g. Standard
English; Miguel argues in
http://groups.yahoo.com/group/cybalist/message/17000 that 'thou'
and 'you' are themselves cognate); 'this' and 'that' ('those' was
originally the plural of 'this'), 'he' and 'they' (sometimes the
same word - nearly so in French) (a foursome altogether); 'who'
and 'what' (are they really separate lexemes in the older IE
languages?); 'man' and 'person' (I'm undecided whether the English
word for 'person' is _man_ or _person_, pl. _people_.).

As I've said before, almost entire numeral systems can be completely
replaced, though IE numerals are remarkably resistant to replacement.

The problem in all these cases is that if one meaning changes its
word, the related meaning is also likely to change its word, thus
effectively reducing the sample size. I don't claim precision in
such matters; my knowledge of statistics has not significantly
progressed beyond undergraduate level, and my teaching of it did not
progress beond a demonstratorship.

Richard.