is in the character of interfaces everywhere in the f-mind to be“dirty”; there is no reason to expect more here.
A theoryofmeaningcompatiblewithgenerativegrammar must bepsychologicalthroughand through:itshouldaspire
to study the nature of thought. The parallelarchitecture, because it is embedded in a larger theory of the f-mind, gives
us a natural way to study the interaction of language with perceptionand to pose questions for research in perception.
It also gives us a way to pose questions about conceptualization in non-linguistic organisms such as babies and apes,
and to ask questions about the innateness and the evolutionary basis of human thought.
The twofinal chapters, surveying issues in word and phrase meaning, give a sense of the complexity of the theory of
meaning. Word meanings are compositional, in some respects paralleling the combinatoriality of phrases but in some
respects proving more complex and moreflexible. Phrasal semantics, like phonology, is organized into a number of
tiers, each of which carries an independent aspect of meaning. Aspects of phrasal meaning, in particular referential
structure and information structure, lead us on into issues of discourse structure, where further complexity awaits us.
A majorbenefit oftheparallelarchitecturehas emerged inthecourse ofthelast chapter: enrichedcompositionand the
multiple tiers of semantics liberate syntactic theory from a great deal of the complexity with which it has become
encumbered over the years. In retrospect, we can see that this complexity has arisen from the demand that syntax be
the sole generative component, responsible for all combinatorial structure in semantics. Now that semantics has its
own generative organization, syntax needs to share with it only enough structure to get words into the right order for
phonology. We can therefore envisiona far leaner syntactic component, taking some of the burden off the learner and
off UG as well.
My intuition is that this is as it should be. Phonology and syntax, after all, evolved toexpressmeaning, which has a far
longerevolutionarypedigree.Moreover, as wehaveseen,muchofconveyedmeaningistacit,notexpressedatallinthe
words uttered or the syntactic structure combining them. Why, therefore, should meaning not be a great deal more
complex than the language that expresses it?
This is not to say syntax will go away altogether. There is still plenty of work for syntacticians. We still have to get the
verb in the right place and supply agreement, case marking, and epentheticit. We have to say which languages have
ditransitive constructions and which don't, which languages require subjects and which don't, how phrasal syntax and
morphology interact, how each language forms questions and relative clauses, and what offbeat constructions it