Re: THEORY: two questions

Tim Smith wrote:

>In fact, I'm inclined to go even further and to say that at this stage,
>_any_ attempt to "explain" language from _either_ a formalist/generativist
>or a functionalist perspective is probably premature, because we don't have
>enough data.  This is _not_ to say that we shouldn't be trying, only that
>real success is still a long way off.  At this point, I tend to agree with
>Dixon (in the book that Matt cited earlier, "The Rise and Fall of
>Languages", which I read a few months ago and was tremendously impressed
>by) that the most important thing for linguists to be doing right now is
>not theorizing but gathering data for future theorists to work on:
>describing as many languages as possible, in as much depth as possible,
>before they go extinct.
Well, I think that data-gathering and theorising need to go hand in hand.
Obviously, you can't have a good theory without good data--but I think
that the reverse of that is also true, that you can't have good data without
a good theory.  Theory and data influence each other.  Data influences
the theory (obviously) by giving the theory something to explain, and by
providing a means for testing that theory.  But theory also influences
data (less obviously) by helping set the priorities for what kinds of
information to gather.  The amount of time for gathering data is finite,
and since our aim in gathering data is to explain how language works
(something which is *only* possible if you engage in theorising), we
need to concentrate our attention on gathering data about those aspects
of language which will help us illuminate how the system works.  Of
course, the problem is massively tricky:  How can determine ahead
of time, and in a completely unbiased way, what kinds of data will be
important?  Ultimately it's impossible.  But that doesn't mean it's
pointless.  And anyway, making judgement calls about what kind of
data to focus on is unavoidable.

Besides, it's impossible to keep theory out of linguistics, even if
you wanted to.  No matter how objective and neutral you try to be,
it's impossible to present natural language data in any sort of coherent
form without adopting some sort of theoretical framework--or at least
a set of working assumptions.  You may steer clear of trees, Optimality
Theory tableaux, and other trappings of "high theory"--but even
assuming something basic, e.g. that the language you're describing has
phonemes and morphological rules, is a theoretical move.

Matt.