Thelogician's paraphrasethe man such that Beethoven likes himprovides a suggestivemodel:thedesignated individual,him,
is in the normal position for satisfying the variable oflike, but it has a special connection tothe man, expressed bysuch
that. Here are a couple of possible notations. (11a) is based on the formal logic notation forlambda-extraction;(11b) is
loosely modeled on the cognitive grammar notation forprofiling(Langacker 1987).^201
In (11a),xis ahound variable(in the logician's sense of binding, not the neuroscientist's sense); the bolded Object-
constituent in (11b) linked to the upper Object has a similar intent. In either case, it is necessary to add a new way to
satisfy an open variable besidesfilling it with content: in (11a),filling it with a bound variable; in (11b), profiling and
externally linking it. I will adopt the (11a) notation, with no prejudice intended.
Bound variable constructions in conceptual structure correspond closely to syntactic constructions such as relative
clauses andwh-questions which involve long-distance dependencies. In some constructions in some languages, the
bound variableis expressed in syntax at a position corresponding to theupperxin (11a), i.e. at the front of theclause.
Such is the case in Englishwh-relative clauses and questions, wherewho/which/what/etc. appears at the beginning;
standard generative grammar calls this“w/fc-movement.” In other constructions in other languages, the bound
variable is expressed at a position corresponding to the lowerxin (11a); this is generally called“wh-in situ,”as in
Chinese and also in English echo-questions such asBeethoven likes WHO?Finally, the bound variablemay be left totally
unexpressed, as in English relative clause constructions such asthe man Beethoven likesandthe man for Beethoven to hire.
The choice among these three possibilities is a syntactic option that has to be
PHRASAL SEMANTICS 385
(^201) Again, I amfinessingmanydetails, in particular howtherelativeclause meaningis attached to thatof thenoun. For instance,some peoplemight want to connectit through
the definiteness feature; I would not be averse to trying this out. In particular, it would more closely parallel the equatives to be discussed shortly.