Logical Gap

[Brian M. Scott]

ECOLING at aol.com wrote:

> Larry Trask writes,
> concerning his criteria for including words as potential early Basque
> monosyllabic words:

>> My criteria are independent of phonological form,
>> and therefore they cannot possibly systematically exclude words of any
>> particular phonological form.

> I have no doubt that Trask believes this.
> But it has been pointed out before that it does not follow.

It does, if one makes an effort to understand what he's
trying to say.

> There is a logical non-sequitur here, which is *one* of the core
> difficulties in the entire discussion we have been having on this subject.

> It is perfectly possible that criteria may be *stated* in words
> none of which have the slightest thing to do with phonological form,
> and yet, in the real world, the combination of those criteria have
> some consequence which end up excluding some category
> of vocabulary elements describable by a particular phonological
> form, at least statistically.

I doubt very much that when Larry says that his criteria
'cannot possibly systematically exclude words of any particular
phonological form', he means to deny this rather obvious fact.
I understand him to mean that any such (statistically) systematic
exclusion will be the necessary result of real correlations
between phonological form and status as an early, native,
monomorphemic Basque word.  And if the phonology of the words
in question really is statistically different from that of
the lexicon as a whole, any criteria that do their job will
of course result in exclusions that are correlated with
phonological form.

> So it can happen, through using a combination of criteria,
> that several strata of vocabulary end up being excluded *statistically*
> (all I have ever claimed) by Trask's criteria.
> Since most of us know independently of Trask's criteria
> that most or all living languages contain vocabulary of such strata,
> it would normally be considered appropriate to include some of them
> as candidates for reconstruction of any pattern which is expected
> to have any sort of general validity for the language.

I don't think that you've ever answered the obvious question:
Precisely how would you choose these candidates?  For that
matter, how would you decide just what strata are inadequately
represented in the written record?

Brian M. Scott