inheritance hierarchies in the brain, only inheritance. My sense is that inheritance is not very well understood formally
yet, particularly in this more general psychological context.
Next we return to the question of why the vast majority of idioms should have perfectly ordinary syntactic structure.
Section6.5 suggestedsomewhatcrypticallythatphrase-structurerulesnotonlyactas vehiclesforfreecombinationbut
also exert pressure for structuring stored lexical items. This suggestion does not make a lot of sense within the
traditional view of phrase structure rules, in which they have nothing whatsoever to do with the lexicon. But in the
present context it suddenly looks more natural. A phrase-structure rule is a lexical item with variables in it; free
combination is achievedby attaching further structure to the variables. At the same time, as we have just seen, phrase-
structure 1-rules such as (35a, b) are sitting in the lexicon, available for other lexical items to inherit properties from
them. A syntactically regular idiom thus inherits a great deal of its structure and is hence simpler than a syntactically
irregular one.
A parallel situation arises in phonology. Principles of syllable structure enable speakers both to syllabify incoming
signals onlineand toconstrain stored words. Inaddition, speakers can use these principles to make distinctions among
putative words, forexampleinHalle's well-knowncontrast betweenthepossibleEnglishwordblickand theimpossible
English wordbnick. And they exert a pressure on the phonological for mof borrowings; for instanceDvorak is
pronounced in English with three syllables in order to avoid the onset /dv/, which is perfectlyfine in the original
Czech but ungrammatical in English. Moreover, just as there are rare idioms with syntactic violations, there are rare
words with ungrammatical syllable structure, for examplekvetch(pronounced /kfeč/in my dialect).
The claim, then, is that principles of free composition act also as constraints on stored items, both in syntax and
phonology. This doublerole fallsout rather naturallyin thepresent approach; itisless natural intraditionalapproaches
(including even Pinker's), where regular rules are entirely distinct fro mthe lexicon.
I a m must ad mit to being uneasy with clai ming that the pressure on lexical ite ms fro mregular 1-rules plus historical
contingency are together enough to account for the overwhelming syntactic regularity of idioms. Historical
contingencies surely are responsible for some /rregular idioms, such asfar be it from me to VP, in the same way as they
are responsible for the irregular and semiregular verbs of English (Pinker 1999). But in order to answer whether
inheritance is enough to account for the regularities, we may have to go beyond linguistics to explore the general
properties of semantic memory. Conversely, evidence from lexical memory can now be brought to bear on the
properties of