an interesting parallel with morphology: apart from their semantics, idioms are like stored regular morphological
combinations. A particularly good analogy might be obligatorily plural nouns. For example, scissors is
morphophonologically regular but does not denote multiple objects (rather an object with multiple parts); and
troops(in the sense‘soldiers’) has no singular for mthat means‘soldier.’These are morphological idioms with regular
morphological form.
Still, the question arises of why there shouldn't be more syntacticallyirregularidioms like (13). One reason is probably
that idioms are historical developments from syntactically compositional phrases. Another possibility, suggested by
Jackendoff (1975), is that the general principles of free composition, which govern online construction, also place
pressure on stored items to conform as well. We will return to this issue in section 6.8.
The second question is whykick the bucketcannot for ma passive or other standard defor mations:the bucket was kicked
by Johnhas onlytheliteralinterpretation.Chomsky has suggested(e.g. 1981) that idioms are somehow“frozen”intheir
deep structure form. For instance, the NP in (10) is inaccessible to passive because it is inside the V. However, such a
solution would preclude moving the PPto taskout of V and around the direct object. Much worse, as pointed out by
Nunberg et al. (1994), many idioms appear only in heavily“transformed”versions. (15) gives some examples.
(15) NP has it made. [passive]
NP isfitto be tied. [passive]
play hard to get [“tough-movement”]
hard to take [“tough-movement”]
a tough nut to crack [“tough-movement”,infinitival relative]
How do you do? [Wh-movement]
(Now) NPe's got NPjwhereproewantsproj. [Wh-movement, free relative]
NPe's not whatproe's cracked up to be. [Wh-movement, free relative]This shows that idioms must be specified in the lexicon for theirsurface structure. The best way to accomplishthis is for
the grammar to check lexical items against surface structures. In a theory withderivational rules (e.g. all of Chomsky's
theories throughout the years), this amounts to what has been called“late lexical insertion”(section 5.7): lexical items
are inserted not at the initial point in syntactic generation, but after all movement has taken place. Of course, in a
theory lacking derivational rules, where surface structure is theonlysyntactic level,lexical insertion/checking/licensing
at surface structure is the only option.
There are, however, idioms that do permit a degree of syntactic freedom, for