416 MARY L. NUNES
More commonly, A's — regardless of the thematic roles associated with
them — are linked to animate NPs which can generally also be human. U's,
on the other hand (with the exception of experiential state U's), are either
linked equally easily to animate or inanimate NPs, or are linked to inani
mate or nonhuman NPs only, a limitation which is rarely placed on the vNP
As included in the data base. These more common linkings are exemplified
below.
(64) a. the editor'sef:A criticism of Mayor Henley/Mayor Henley's
decisionth."U"
b. Jess'sef:"U" avoidance of the problem/Sarath:"U"
c. John'sef:A concealment of the knife/the childpat:U
d. Joe'sloc:A inheritance of the money/the Arabian horsesth:U /
??Jane (human referent)
e. Bob's ag:A invention of the echo battery/"Aunt Sue" (imaginary
character)Pat:U/*Sue (human referent)/*the lynx
f. Jcin'sâg.A amendment of the book pat:U]/*her son/*the puppy.
In other words, in English, animacy and human status are more typi
cally and more freely associated with the semantic macrorole of A than U.
Consequently, in vNPs including an U direct argument whose thematic
relation does not by definition reflect animacy (= EXP), the A is permitted
to function topically in the LDP.
Where an U direct argument is not included in the vNP, however,
other variables interact with animacy to permit argument preposing. Thus,
As cannot always function topically, even if they are linked to a lexicalized
agent (associated with by definition animate — and therefore inherently
topical — NPs) and even if an EXP is not included in the vN's LS. The
ACM assassination, for example, takes an AG A and a PAT U. In Ken
nedy's assassination, Kennedy can only be interpreted as the PAT U
affected by the action, not as the AG A responsible for the action. Since
the thematic relation of PAT need not have an animate referent (e.g. the
city'spat destruction), affectedness obviously takes precedence over animacy
in permitting preposing in nominals lacking a direct argument and headed
by vNs like assassination and destruction.
Affectedness is, of course, the criterion Anderson (1979) used to
defend her analysis of "object preposing." Lacking theoretical principles
which could be applied to the notion of affectedness, however, she was
forced to rely on inuitively-drawn definitions which addressed only specific
sets of preposed "objects." In distinction to the configurational framework