IE and Uralic pronouns

[Larry Trask]

On Sun, 7 Mar 1999 ECOLING at aol.com wrote:

[on pronoun comparisons]

> Rather appear to be knowing the answers,
> and hide the questions we don't know the answers to?
> Or declare them off-limits?

Nobody I know of is doing any such thing.  We merely admit that there
are many things we don't know, and quite a few things we are unlikely
ever to know.  But no historical linguist declares any field of
investigation "off limits".

> In a business firm, that would be a recipe
> for near-immediate bankruptcy,
> failing to act positively and strongly in moving towards the future.

This is the first time I have seen HL compared to a business.

HL is not a commercial enterprise.  It cannot be run like a business or
judged like a business.

Our late unlamented Conservative government in Britain attempted to
introduce business procedures into such institutions as schools,
hospitals, universities. the police and the BBC, all of which were
burdened with layers of managers and bureaucrats, and assessed as
business enterprises instead of as public-service institutions.
The results have generally been catastrophic.  And I don't think
business attitudes would be any more rewarding in my subject.

> Better to ADMIT to students that we need more powerful
> tools for penetrating the noise of centuries and even millennia,
> to set more and more students to this task, to studying how deeply,
> and against which kinds of noise, each tool we have can
> penetrate (estimates in all cases are fine), and to estimate
> in as many situations as possible how much residue should
> still be detectable despite the noise of historical change,
> WITHOUT the assumptions that the sample used as controls
> ARE IN FACT UNRELATED.  That is the fallacy in all such
> approaches I have read.  Because if the supposed controls
> ARE related, even very distantly, it may distort our estimates
> of what background noise is.

With respect, this sounds to me like a hopelessly unrealistic program.
A number of linguists are already attempting to develop mathematical
methods which might be used to push our investigations further back in
time, but the difficulties are formidable, and no proposal has yet won
widespread acceptance.

I don't think it's realistic to try to estimate the degree of background
noise for languages generally: there are just too many complicating
factors.  For one thing, languages with similar phoneme systems, similar
phonotactic patterns and similar morpheme-structure constraints are
likely to show a higher proportion of chance resemblances than arbitrary
languages.  For another, no general approach to background noise can
hope to distinguish between common inheritance and borrowing; this is
something that has to be done by painstaking and hard-nosed linguistic
investigation, and sometimes it can hardly be done at all.

As for the supposed "fallacy", I might draw attention to the Oswalt
shift test, a simple but ingenious -- though time-consuming -- way of
estimating the degree of background noise for any languages we happen to
be interested in.  After all, if we're interested in comparing, say,
Burushaski and Ainu, it's only the background noise for those languages
that is relevant, and not the background noise for other languages,
which may be quite different in nature and magnitude.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk