The syntax in this architecture determines a set of syntactic categories which are domains of word and phrase order
regularities. For instance, the word order possibilities within a phrase are the same no matter where the phrase occurs
in the sentence. In addition, syntactic phrases are domains for interface principles to semantics; for instance, Agent
First is now an interface principle between nounphrasesand thematic roles rather than betweennounsand thematic
roles.
At the same time, because evolution is incremental, the direct relation between phonology and meaning does not go
away. In particular, the connection between word meanings and word pronunciations is direct, bypassing syntax. And
thisimportantcharacteristicremains inmodern language: as has been observedseveral times already, syntacticfeatures
of words are far coarser than phonological and semantic features. For instance, as far as syntax is concerned,dog, cat,
worm, amoeba, andtreeare indistinguishable: they are all just singular count nouns. Similarly, verbs with identical
argumentstructuresuch asjog, walk, shuffleandstrutare syntacticallyindistinguishable.Thus thesyntacticfeatures serve
only to tell the syntax how this word canfitintoa phrase structure; they correspond only very coarsely to meaning. (As
Levelt 1999 puts it,“Syntax is the poor man's semantics.”See also Pinker 1989, where the semantic coarseness of
argument structure plays an important role in the theory of children's acquisition of verbs.) The mapping between
phonology and fine-scale meaning differences ought therefore to be maintained as part of a straight
phonology–semantics interface without syntactic intervention. (And the“defective”words likeouch haveonlythis
connection.)
8.10 Grammatical categories and the“basic body plan”of syntax
Sofar wehavemanaged todo without thedistinctionbetweennouns and verbs. Everythingis donesemantically. How
does this grammatical distinction arise along with the further differentiation of adjectives, prepositions, and so on? I
have only a very speculative story here, but I will offer it anyway.
An important asymmetry between nouns and verbs came up in section5.5. Nouns can express any semantic category
whatsoever: notjust objectsbut situations (situation, concert, earthquake, perusal), properties(size, intelligence, redness), spatial
concepts (distance, region, place), times (Tuesday, millennium), and so on. But verbs can express only situations (events,
actions and states). In section 5.5 this was left as an unexplained design feature of the syntax–semantics interface.
Carstairs-McCarthy (1999) observes another significant asymmetry in language, one that is not at all necessary to its
expressive power. The sentence (4a)