approach is that the idea of multiple generative components connected by interface components has been extended in
thoroughgoing fashion to every part of the grammar, so that it becomes a fundamental architectural design principle.
A similar overall plan emerges in the musical grammar proposed by Lerdahl and Jackendoff (1983). There the
generative components aregrouping structure, metrical structure, time-span reduction, andprolongational reduction.Thefirst two
of these, together dealing with rhythm,find close parallels in the prosodic syste mof language. The last two, dealing
with melodic and harmonic organization, bear no resemblance to anything in language.
More generally, it is abundantly clear that the brain is organized in terms of numerous interacting areas that together
determine our experience of the world and our intentions to act. A description of the f-mind in terms of independent
components that interact through interfaces thus offers the hope of an interesting account of its instantiation in brain
terms. As observed in Chapter 2, such a possibility is of course a desideratum, even if the connection between
functional and neural theories is far in the future.
5.7 The lexicon and lexical licensing
We next turn to the lexicon. For afirst approximation, the lexicon is the store of words in long-term memory from
which the grammar constructs phrases and sentences.
It is widely agreed that a word is to be regarded as a long-ter m me mory association of phonological, syntactic, and
semantic features. Many have suggested that words that denote physical objects also carry with them some sort of
image of a stereotypical instance; in Fig. 1.1 this is notated as the spatial structure associated with the sentence. If a
wordisconcernedwithanothersensorymodality, for instancethewordacrid, itwillpresumablybelinkedtoa structure
appropriate to that modality instead. We return to spatial structure and other sensory structures in Chapters 9 and 11;
for the moment we will consider only the phonological, syntactic, and conceptual features associated with words.
Recall that mainstream generative grammar, following Chomsky (1965), inserts lexical items as a whole into syntactic
structure; their phonological and semantic features are interpreted later in the derivation by the appropriate
components. While this approach is not formally impossible, it raises the question of why syntax should drag around
all these features that it cannot itself access. As Ivan Sag has suggested (p.c), it is as though the syntax has to carry
around two locked suitcases that it turns over at a checkpoint to the components that have the right keys.
A number of people over the years have noticed this anomaly and suggested a process of“late lexical insertion”(e.g.
Otero 1983; den Besten 1977; Fiengo