concentrated more on interface components than generative components here, it is because in my opinion they have
not previously received enough attention. People speak of “the phonologysyntax interface” but not “the
phonology–syntax interfacecomponent,”failing to recognize interface rules as qualitatively different fro mgenerative
and derivational rules.
The move that most decisivelysets the present framework apart from mainstream generativegrammar is its treatment
of the lexicon as an essential part of the interface components. A word likecatis not a list of phonological, syntactic,
and semanticfeatures that is inserted intosyntax and carried around in the course of a derivation.Rather, it is a small-
scale interface rule that helps correlate the parallelstructures. Its phonology appears only in phonological structure, its
syntax only in syntactic structure, and its semantics only in conceptual structure; and the word as a whole establishes
the linking among the three in the overall structure of a sentence. Here we concur with some of the constraint-based
frameworks, especially HPSG.
This sets the stage for a second step in reconceptualizing the lexicon.The notion of lexicon traditionally conflates two
very different notions (at least informally): a list of words and a list of things that are memorized. The foundational
observation of generative linguistics, of course, is that productive combinatoriality cannot in general be the
consequence of memorization, but must be composed online by applying rules. By taking this dictum very seriously,
we are led to recognize that lexical items include not just words but also productivemorphological affixes and idioms.
The treatment of lexical items as interface rules makes this move not only plausible but conceptually and formally
natural. From idioms itis a smallstep torecognizeidiomaticallymeaningfulconstructions as lexical itemsofa newand
previously unrecognized sort—more or less as advocated by Construction Grammar, but in my opinion better
integrated intothesystemas a whole. Andfinally, thisputs us ina positiontorecognizesuch things as phrase structure
rules as lexical items consisting entirely of variables. Quite possibly the only combinatorial or “procedural”rule
remaining in the grammar, then, is UNIFY.
Fro mthe point of view of traditional generative gra m mar, we have by this point co me to live in a very strange and
perhaps uncomfortable land. I cannot myself claim to have come to terms with it entirely. However, I have tried to be
as clear as I can aboutwheretraditionalissues insyntax, phonology, and morphologyfit intothenewsystem. My hope
is that the large-scale map I have offered may motivateexploration of more of the details, especially at the boundaries
where major portions of the map have to match up. The treatment of argument structure in Chapter 5, of VP
constructions in Chapter 6, and of lexical and phrasal semantics in Chapters 11 and 12 were offered as samples, these
being topics on which I personally have spent a lot of effort over the years.