Hallo!

On Sun, 25 Jul 2010 15:59:16 +0100, David McCann wrote:

>  Basically, I don't expect to convince you, or Ray. I work on the basis
>  that all languages are ultimately related, since we are all descended
>  from one woman.

While I consider it *likely* that all languages are ultimately related,
it is still *unproven*, and presupposing that they are is meaningless.
You begin with a result and try to find evidence for it; that is the
way crackpots work.  Real scholars start with the available evidence
and see what kind of results can be obtained.

>        The task is tracing the relationships we know must
>  exist.

No; the task is finding out *which* relationships exist.  There are
*no* relationships (apart, of course, from those already established
by means of the comparative method) which "we know must exist".

>        Such work has to take geography, prehistory, and genetics into
>  account.

These factors may be useful *hints* at what happened; but people move
around a lot, so geographically distant languages may be related and
neighbouring ones may be less so; and language shift (i.e., a
population switching from one language to another) happens often
enough to invalidate all attempts to equate genetic and linguistic
relationships.  What counts in historical linguistics are the
*linguistic* data, and the only means to show that two languages are
related is the comparative method, which operates on regular sound
correspondences.

>        Nostratic is plausible on that basis, and evidence for it only
>  confirms what one would expect.

Nostratic may be plausible, but "plausible" doesn't mean "proven".
The evidence I have seen so far for it is far from convincing, and
most scholars think so as well.  There are numerous problems with
it: irregular sound correspondences, implausible semantic connections,
recourse to data from individual languages which don't take the known
histories of the families concerned into account (e.g., words which
probably aren't inherited from the protolanguage).


>        I don't compare Latin and Swahili, Greek
>  and Malay (as poor old Swadesh was prone to do) because any connection
>  would take one back to the emergence of Homo sapiens from Africa, too
>  far back for anything to be reconstructed. If you want to conduct
>  linguistics in a vacuum, or with arbitrary methodological constraints,
>  you will naturally get different results.

The mainstream academic scholars do not "conduct linguistics in a
vacuum"; they work with *the available evidence* and with *as few
assumptions as possible*.  You start with a strong assumption, which is
not proper scientific methodology.  The methodological constraints of
the academic mainstream are *not* arbitrary; they are based on
experience of past scholars which methods yield good results and which
do not.  If you look at what has been published about Nostratic, you
will see that there are two major (and several minor) camps who posit
different sound correspondences and each manage to find several hundred
"cognate sets".  The two reconstructions are incompatible with each
other; they cannot both be right.  Clearly, a method which yields so
many false positives is not reliable.

>        They will, however, not
>  convince me. I suspect dialogue between us is impossible.

I suspect so, too.  Learn about how historical linguistics works.
There are several good and easy-to-read textbooks, such as that by
Lyle Campbell (titled simply _Historical Linguistics_).  It should
not be difficult to find a library which has it.  If you refuse to
accept proper methodology, you will fail forever.

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html