Default unification: online vs. offline

[Luis Casillas]

On Mon, Oct 22, 2001 at 08:11:37PM -0400, Robert Levine wrote:

> *Is* there either data or psycholinguistic evidence which actually
> point unequivocally to defaults of any kind as anything other than an
> abbreviatory device?

Well, my question is partially posed in face of other theories of
grammar that take as a starting point that there is such data, e.g.
Langacker's Cognitive Grammar, where "categorization", which is taken
to be one of the basic cognitive processes on which grammar is based,
works by means of classification of objects in terms of prototypes. The
objects categorized by a prototype don't have to show all the properties
of the latter. Well entrenched, automated categorizations are a rough
analogue of the HPSG type hierarchy, while spontaneous categorization
when faced with actual utterances is the analogue to processing.

The application of this idea in Cognitive Grammar in some places bears
resemblance to the use of default constraints in HPSG; e.g. the CG
equivalent of "supertype" is "prototype", and similarly, "subtype"
corresponds to "prototype extension", and given that the extension
doesn't have to fully match to the prototype, there is some analogue of
"overriding defaults" going on.  And this is allowed to happen in online
("nonautomated") processing.

So there certainly are people out there who believe that the answer to
your question is "Yes," and build their theories of grammar accordingly.
My answer is "I don't know, that's why I'm asking..."

--
Luis Casillas
Department of Linguistics
Stanford University