Chomsky and deep structure

[Brian MacWhinney]

David and FunkNet,

   Not to distract from the really fascinating issues being discussed in the
functionalism-and-OT thread and the corpus linguistics thread, I thought I
would take a shot at a quick response to David Tuggy's reasonable request for
a statement regarding *the* most problematic of Chomsky's positions.

   As far as I can see, the core problem with Chomsky's generative grammar is
the commitment to deep structure.  All of the other problems derive from that
commitment.  "Underlying form" is a reflex of deep structure in that it
allows us to map the many onto the few, thereby appearing to achieve
"linguistically significant generalizations" along with abstractions that can
be used for further abstract theory building.  "Competence-performance" and
the "ideal native speaker" are constructs that are required to support
theories of deep structure.  Without deep structure, the notion of universal
grammar is hard to imagine.

  I realize that the importance of deep structure has waxed and waned in the
various versions of the theory.  But I see no evidence that, the core
conceptual importance of deep or abstract structure has ever disappeared.

  What is wrong with deep structure?  The problems have to do with the (1)
the nature of empirical evidence and (2) the structure of human thought.
People don't speak deep structures and it is essentially impossible to find
them in natural interactions or experiments.  Of course, one can argue that
physicists rely on a particle and force-based deep structure in quantum
dynamics.  This is true, but the hypothesization of particles has been
supported at each point by precise empirical measurements matched with
detailed mathematical accounts that are remarkably precise.  No such accuracy
of measurement supports the constructs of linguistic deep structure.
Psychologists like Sigmund Freud have shown us how dangerous it can be to
construct theories of deep structure that cannot be clearly pegged to
empirical tests.   I am certainly not opposed to abstract theories in
principle, but I am only willing to work with such theories when the
individual components of the theory have demonstrable empirical grounding.
Many of the constraints of certain versions of OT seem to fulfill this
requirement, for example.

  My second reservation about deep structure has to do with what we know
about other areas of perception and cognition.  It is certainly true that the
construction of a visual percept relies on low-level feature detection.  In
this sense, we could think of vision as building up constructs from a deep
structure base.  But this is not what Chomsky has in mind with the
non-perceptually-grounded categories of universal grammar.  In principle, one
could imagine that a cognitive system could have arisen through evolution
that was structured in this way.  But the fact that language is a relatively
recent event in primate evolution makes me wonder how something so abstract
and ungrounded as deep structure could have arisen in such a short
evolutionary time span.  Calling language a sprandrel (ala Chomsky and Gould)
is fine, but sprandrels are simple emergent structures and deep structure is
not.  So where would deep structure come from?

  One can argue that deep structure is just a hypothesis that should be
pursued with the same tenacity with which we pursue any scientific
hypothesis.  I agree with that.  But, just as I would not place all my
funding resources on one approach to fusion research, I don't think that all
of linguistics should be dedicated to pursuing theories grounded on deep
structure.  In reality, linguistics is far more diverse than the general
public realizes.  Some high profile media events have made it seem
monolithic, but this is misleading.  I think that Chomskyan approaches will
never disappear, since the hypothesis of deep structure will always be
tenable and worth exploring.  Once we really understand how language is
processed and stored by the brain, we may reinvent a empirically-grounded
type of deep structure.   But it will have only a vague family resemblance to
the deep structures of "Cartesian linguistics" or Chomskyan linguistics.

  What is right about Chomsky?  I think Chomsky's great contributions are his
emphasis on generativity and mechanism, his treatment of language as a
component of human cognition, his ideas about language and creativity, and
his stimulating collaborations with George Miller and Eric Lenneberg.

  So, David, that's how I see it.

--Brian MacWhinney