However, his notionof definition is the standard dictionary sort: a phrase that elucidates a word meaning. So what he
has actuallyshown is thatword meaningscannotbe builtby combining other word meanings, using the principles that
also combine words into phrases.^172 But there are other options.
A comparison with phonology is useful. The phonology of a word is certainly not monadic and innate; it is built from
known parts. At thefirst layer of decomposition, wefind a repertoire of language-specific syllables; at the second, a
repertoire of partlylanguage-specific speechsounds; at thethird, a universalrepertoireof distinctivefeatures. What we
take to be innate is the repertoire of features, plus general principles for building the minto language-specific
repertoires of sounds, syllables, and words.
Notice what happens as we decompose further and further. Among the syllables of a word, many could be words on
theirown (e.g. withinsyllables, sillandbulls). At thenextlayer down, most speechsounds cannotbe wordson theirown
(vowelscan, butindividual consonantscanonly beclitics, as incats, he'll, I'm, named)—but at leasttheycan beintuitively
discriminated. However, at the next layer down, distinctive features not only can't be words on their own, they can't
even besoundson their own. We have no conscious access to them, and it requires a theory of phonology to uncover
their differentiation and their role in building words. Moreover, although speech sounds combine into syllables by
concatenation, and syllables combine into words by concatenation,and words combine into phrases by concatenation,
distinctive features do not combine into speech sounds by concatenation.
Whether or not we suffer physics envy, a physical parallel is also suggestive. In explaining the variety of substances in
the world, afirst layer of decomposition gives us a limited repertoire of substances such as oxygen, boron, sulfur, and
gallium, combined by chemical bonds. The next layer down gives us entities that can exist in isolation but are not
substances: electrons, protons, and neutrons, combined by electromagnetic and nuclear forces. The next layer down
gives us quarks, which not only are not substances but cannot exist in isolation; they are as it were features of
elementary particles.
I suggest that the same is true of lexical concepts. As we explore the parts from which they are built, suppose wefind
layers of structure whose units cannot individually serve as possible word meanings. This would preclude an
explication of
LEXICAL SEMANTICS 335
(^172) For more detailedcritiqueofFodor's position, see(Jackendoff1983 : 122–7; 1990a : 37–41, 150–2). As far as I a maware, Fodor has notrepliedto thesecritiques. He does
offer an argument against my type of decomposition in Fodor (1998 : 49–56) (also discussed in Laurence and Margolis 1999 : 58–9). But his argument is made only on
methodological grounds and completelyignores all empirical generalizations captured by the decompositional approach. We return to this argumentbriefly in section 11.8.