formal/functional: a commercial

[Dan Everett]

Responding partially to Dick Hudson, partially to myself:

There is a serious problem, however, with formal accounts, especially
Chomskyan, though, which renders the issue of what to study a bit more
serious than 'where to put our research effort', as Dick puts it. This
will not be news to many readers of this list, but the light has just
come on for me, so let me say how it did.

The problem is just this: Formal linguistics, at least of the straight
Chomskyan pedigree, has no intensionally definable object of study. It
claims to study "I-Language". But what is I-Language, really? It is just
grammar. But what kind of grammar? It is the grammar of an individual
speaker, abstracting away from historical, functional, and sociological
properties of that grammar. However, how can we know, a priori, what
aspects of the grammar fall into these three categories? The answer is,
we cannot. Therefore, one cannot define the object of the study of
I-Language grammars a priori. There are no properties which can pick it
out except, perhaps (and I doubt this for most cases) empirical
properties. But if this is so, then I-Language can only be defined
extensionally. What would that mean? It would mean that I-Language must
be reduced to the grammar of a set of sentences. Which? Those the
researcher chooses to study. What happens if these sentences turn out to
be products of functional, historical, or sociological forces? Move on
to another set of sentences.

I have ceased to be sanguine about the possibility of building an
interesting research program on such an elusive object of study.

Functionalists do a little bit better. They are willing to study
everything related to language. One can give an intensional definition
of this. But it would be a bit, uh, broad.

Neither of these two approaches has a lot to cheer about when it comes
to define what it is studying. But I think functionalism does better.

In any case, although I still am working on some formalist phonology,
this is why, more and more I have come to believe that the best activity
for the linguist to engage in is writing grammars and *from these
efforts* offering 'explanations' of what we find important, puzzling, or
just interesting. And this is why I have come to be convinced, finally,
that any serious theory of grammar must allow non-formal information to
be part of its architecture. Semantics and pragmatics must be causally
implicated in the theoretical constructs of a good theory of grammar.
Yet, at the same time, formalist discoveries must be accounted for. The
most important empirical findings of formal grammar were made in 1967 by
Ross, following a 1964 lead of Chomsky's: the island-constraints.

Any theory of grammar ought to be able to handle these, yet avoid the
problems I have just mentioned about the object of study. I only know of
one theory of grammar that meets this requirement, RRG. So it is my
growing disquiet with the disappearing object of study of formal
grammars, what I call in a paper in progress, "The shrinking Chomskyan
corner", coupled with my concern for accounting for the semantics,
structure, and pragmatics of natural language that have led me to RRG as
the best port in the storm.

On the other hand, none of this changes my belief that we ought to all
be asking for NSF money to get into the field and describe languages.

End of commercial.

What I would like to hear (and this isn't, I know, the best list on
which to pose the question) is if anyone can think of how one might go
about defining the object of study for a formal theory.

Dan Everett