This is somewhat more elaborate thanstaranddevour(Ch. 5, (24) and (51a)), but the similarities are clear. In particular,
theargumentstructureis justlikethatof a normal verb: thetwoargumentsin conceptual structuremust beexpressed,
justas withdevour; oneisexpressed as thesubjectand theother as thedirectobject.However, (14)has twoinnovations.
Thefirst is that the phonological and syntactic components of the item are larger than phonological and syntactic
words respectively. Second, the subscripts are now bifurcated into pre-subscripts, which denote phonology–syntax
connections, and post-subscripts, which denote syntax–semantics connections. This is necessary because an idiom
doesnothavea standard word-by-wordmapping. Here,thefirstwordinphonologymapsintotheverb, thecliticmaps
into the preposition, and the second word maps into the noun. But these individual words do not have meanings of
their own: only the whole verb phrase does—it means“criticize,”as notated by the subscriptm. That is just what
makes (14) an idiom.
The lexical structure in (14) permits a simple account of the idiom's discontinuity. The idiom's second semantic
argument maps into an NP in syntax, perhaps automatically, perhaps by strict subcategorization (I leave the question
open). Because of the Syntactic Argument Ordering Constraint (Ch. 5, (47)) this NP must precede the PP that is
inherentlypart of the idiom. In turn, the corresponding phonological material falls betweentakeandto taskbecause of
the NP's syntactic position.
The non-homogeneous subscripting across components in (14) bears an interesting parallel to irregular verb forms
such as ate (7). There we found a regular compositional syntax–semantics connnection but an irregular
phonology–syntax connection: the syntactic and semantic units cannot be discriminated in the phonological structure.
Idioms like (14) have a regular phonology–syntax connectionbut an irregular syntax–semantics connection: the words
do not contribute individually to the meaning. Thus the two cases are in a way mirror images.
Three important questions are always posed of theories of idioms, and they must be addressed here. Thefirst is why
virtually all idioms have perfectly normal syntactic structure, with only rare exceptions such as (13). I would propose