instanceHe let the cat out of the bagvs.The cat was let out of the bag; We must draw the line somewherevs.The line must be drawn
somewhere. Ithas frequentlybeen observed (Bresnan 1978; Ruwet1991; Nunbergetal. 1994) thatthemobileportionof
such idioms has a sort of metaphoricalsemanticinterpretation:the catis a secret,the lineis a distinction.By contrast,the
bucketinkick the buckethas no such interpretation. The key to mobility appears to lie in the partially compositional
semantics of the idiom. A proper solution is beyond the scope of this chapter (but see Jackendoff 1997a: 166–70 for a
proposal).
The third question about idioms is whether a structure like (14) might be too specific. Perhaps, for example, one
should remove allindicationof syntactic order, since itis predictablefrom the phrase structurerules in any event.And
perhaps one should replace the phonological content with pointers to the lexical itemstake, to, andtask, which would
provide the idio mwith its phonology“on the cheap.”
I wouldsee such movesas reflections of theimpulsetomake thelexiconconsistonly ofnon-redundant, unpredictable
information. We have already implicitly rejected this position. Consider stored words with regular morphology: they
can be generated by free combination, so they are totally redundant. What would it mean to remove all the redundant
informationfrom storedregulars? Itwouldamounttoderivingthembyrulesalone—whichisexactlywhatisdeniedin
claiming they are stored! The case of idioms is similar. Parallel to the predictable order of stem and affixes in stored
regulars is the predictable word order of idioms. Parallel to the predictable phonological content of stored regulars,
which comes directly from the stored stem and affix, is the predictable phonological content of idioms, which comes
fro mthe constituent words. I conclude that if such redundancy is tolerable in stored regulars, it should be no proble m
in idioms. We return to this issue in section 6.8.
6.6 A class of constructional idioms
I now want to move from idioms like those in (11) to a broader class of lexically specified VPs that include free
variable positions. All the members of this broader class have the syntactic structure shown in (16).
They vary in which parts of the VP are open positions. The idioms in (11) are a subclass that has a specified V and a
specified PP but leaves theNPopen. Wecan schematizethissubclass as (17), notatingthespecified idiomaticelements
in