(65) a.BILL took SALLY to school, and FRANK took PHIL there.
b. BILL took Sally to SCHOOL, and FRANK took her to the MOVIES.Each clause in (65) contains paired contrastive foci, each of which is clearly a constituent. On the other hand, the rest
seems to be a nonconstituent:take...to schoolin (65a) andtake Sally...in (65b).^223 So it looks on the surface as though
Common Ground cannot be coindexed with anything.
A solution arises from theobservation (Chomsky 1972a; Jackendoff 1972; Akmajian 1973) that theCommon Ground
is actually something likesomeone took someone to schoolin (65a) andsomeone took Sally somewherein (65b). That is, the
Common Ground is the result of replacing the Focus and First Focus constituents with variables. In the present
framework,theproper placetodo thisis notin syntax, but inthedescriptivetier—whereit turns outthat thevariables
are already there! They are the open variables that serve as attachment points for the Focus and First Focus
constituents. (66) illustrates. The dashed lines enclose the pieces of the descriptive tier that correspond to the pieces of
information structure, overlapping at the attachment points.
In other words, we can think of the foregrounded elements Topic, First Focus, and Focus as“carving pieces out”of
the descriptive tier, leaving behind a Common Ground with variables in it.
This approach to theCommon Ground yields an elegant account of certainkinds of ellipsis. Consider the examples in
(67).
(67) a.JOHN likes PIZZA, and BILL(,) BANANAS. [Gapping]
b. JOHN threw his BOOK in the garbage, and FRED did the same thing with his TERM paper. [Do the same
ellipsis]PHRASAL SEMANTICS 415
(^223) There have been proposals in the literature that make one or the other of these into a syntactic constituent. Larson (1988) , through heavy-duty invocation of syntactic
movement,evenmanages tohaveitbothways. ButI thinktheproperconclusionis thatneitheris a constituentand thattheVP has theflattripartitestructureV-NP-PP. See
Culicover (2000).