feature structures vs. feature structure descriptions

[Carl Pollard]

> Many points of view on the formal status of features structures and
> their descriptions do indeed seem possible, and I find it interesting
> to see different views colliding in this discussion concerning
> questions that have also puzzled me. I wonder what you mean by saying
> "feature structures are models of linguistic expressions". As a
> model-theoretic semanticist, I am familiar with the notion of a model
> of a language: the mathematical object (M), typically a set-theoretic
> structure, containing the elements that serve as semantic
> interpretations of the expressions of a language (L). My impression is
> that Lutz Gunkel is thinking in these model-theoretic terms of
> languages and models, whereas your notion of model of a linguistic
> expression seems to be different; could you say a bit more about that?

Hi Harry,

I'm actually using the term "model" two different ways at the same
time: the model-theoretic semantics way, and the Poincare/Mach/
Einstein/Hadamard/Hilbert philosphy-of-science way, in which
an empirical domain is theoretically represented by a mathematical
model, with the two linked by a conventional interpretation
that relates observables of the empirical domain to aspects of the
model. For example, Poincare modelled systems of n bodies moving
under mutual gravitation as vector fields on differentiable manifolds.
But the set of differential equations to which the vectorfield is
a solution, if formalized (say in ZF) is at the same time a first-order
theory with the vectorfield as a (semantic) model. In the linguistic
analogy, English corresponds to (say) the Earth-Sun-Moon dynamical
system, the grammar to the system of differential equations, and
a certain set of abstract feature structures to the vectorfield.

Carl