Of course we were all taught as graduate students that syntactic derivations don't model the course of processing.
Speakers don't think of the initial symbolSfirst, then gradually expand it till they choose the words, then push the
pieces around untiltheyfinallyarriveatwhatthesentence means and howtopronounceit.Sothetheoretical notionof
derivation, particularly syntactic“movement,”has been abstracted away fro mprocessing by calling it“metaphorical”.
The derivation is part of competence but has little to do with performance.
This stance is necessary because the syntactocentric architecture has a logical directionality: it begins with syntactic
phrase construction and lexical insertion, and branches outward to phonology and semantics (Fig. 7.1a). This is quite
at odds withthelogical directionalityof processing, wherespeechperceptionhas to get fro msounds to meanings(Fig.
7.1b) and speech production has to get fro m meanings to sounds (Fig. 7.1c).
Linguistic theory can rhetorically make a virtue of necessity by saying“But of course we're developing a theory of
competence; we're interested in the character of abstract knowledge.”Still, other things being equal, a theory that allows
us readilytorelate competencetoperformance oughttobefavoredoveronethat createshard boundaries betweenthe
two.
Let me make clearer what I think is at stake. A theory of processing is concerned with how a language user, in real
time, creates structures for perceived and produced sentences. And a theory of competence must at the very least
account for the range of structures available to the language user. But I now
Fig. 7.1 Logical directionality in competence and performance