Excluding data

[Larry Trask]

On Fri, 17 Sep 1999, Jon Patrick wrote:

[on my comments on Basque <taup> `sound of a heartbeat']

> This comment is a red herring. My commentaries were not about the
> inclusion or exclusion of this word in the analysis but that your
> criteria have high correlation with a model of the phonology that
> you object to being re-analysed from a different perspective.

Jon, I am baffled by your persistence on this point.

I have made it perfectly clear that my criteria for selecting words are
non-phonological in nature, as indeed they must be.  I cannot possibly
select words according to predetermined phonological criteria in order
to make a study of the phonological properties of the result: that would
be pointless.

I exclude <taup> because of its very late first attestation and because
of its limited distribution in the language.  That the word has such a
curious phonological form does not surprise me, but its form has nothing
to do with its exclusion.

[LT]

>     Well, I have nothing against computers, but I don't regard them as a
>     form of magic.  If you put together a list of words 95% of which are not
>     ancient in Basque, what exactly can your computer program do that will
>     be informative about Pre-Basque?

> Once again you are choosing to make inaccurate assertions. I think it was
> discussed previously in the list as the Straw Man arguement.
> A counter arguement to your position is "if you put together a list of of
> words of 5% of which are only ancient Basque how informative of the whole
> picture is that."

But I'm not interested in "the whole picture".  I'm only interested in
the forms of Pre-Basque words, and the vast majority of modern Basque
words did not exist in Pre-Basque.

Anyway, we have no words which are "only ancient Basque".  If a word is
not recorded in the historical period, then it simply does not exist, as
far as we are concerned, and it is not available for examination.

[LT]

>     But the generalizations can come only after the list has been compiled
>     in the first place.  We are talking about how the list should be
>     compiled, not about the generalizations that will emerge from it --
>     though, as I have pointed out often, I *think* I have a pretty good idea
>     what those generalizations will look like -- though I'm prepared to be
>     surprised on occasion.  But, once more: I *never* exclude a word from my
>     list because it doesn't match any generalizations about form which I may
>     have in mind.

> I've never asserted that you did. However I do think that your
> criteria are designed to create an analysis that is more strongly
> consistent with the generalisations you "think you have a pretty
> good idea" about.

I flatly deny this, and I challenge you to back up your assertion.

> My comment is that the human mind is more frail
> than we give it credit and a computer based analaysis helps us be
> more rigorous in our undertakings. It is unwarranted that you imply
> I present it as a tool of magic.

What I'm objecting to is your apparent suggestion that taking the words
off the page and putting them into a database on a computer somehow
frees us from the necessity of choosing criteria for selecting words for
whatever purpose we have in mind.  Databases are easier to work with
than paper, but they are useless until we choose to do something with
them, and that means selection according to criteria determined by the
human investigator.

> I think we have seen an example of
> the extensiveness if not rigour of method that the computer can
> assist us with from the small analysis of the consonant cluster
> <-ltz> I presented in the previous message.

I have already agreed that it is helpful to have a convenient and fast
way of answering questions like "what attested Basque words end in
<-ltz>?"  But the result obtained is meaningless until it is
interpreted.  In this case, it is trivial to show that, of the four
words turned up, only <beltz> `black' can plausibly be projected back to
an early stage in a form exhibiting this final cluster -- exactly the
conclusion I had already reached without using a database.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk