Comparative Methodology

[LV Hayes]

----------------------------Original message----------------------------
>Date:         Mon, 6 May 2002 21:33:34 EDT
>From:         Rich Alderson <alderson+mail at panix.com>
>
>>I was thinking initially of the simplest case where 2
>>languages are compared in order to reconstruct the proto-
>>language of which the 2 are direct descendants.  This
>>scenario would provide the initial guidelines to establishing
>>the requisite phases and steps.  Then, more complex
>>situations could be addressed.
>
>I think you will find that the majority of workers in
>comparative methodology do not do binary comparisons;
>Bloomfield, for example, felt that 4 languages made up the best
>size sample for CM.  Binary comparisons do not give you any
>check on reasonableness of the growing reconstruction as you
>proceed, indeed may mislead you altogether.

I wonder if you're thinking of the principle of triangularity.  I forget
who came up with this, but it argues that 3 comparisons are better than 2
because 2 correlations may be due to chance convergence.  I thought this
principle to be applicable only to compared data, but perhaps it is also to
languages, in which case it would be best to have 3 or more languages to
compare.  Of course, one may not always have 3 or more to compare.

In any case, I was not advocating a procedural format geared solely to
binary comparisons.  Such a format should be context sensitive and
adaptive.  Its development would start with a binary comparison as the
simplest case and evolve as additional cases with more complex situations
are evaluated.

LV Hayes
lvhayes at worldnet.att.net