ARGUMENT LINKING IN DERIVED NOMINALS 413Type A nominals in Table 1 comprise a group of vNs which present a
particular problem in analyses making the double genitive assumption and
attempting to define syntactic relations in the vNP in terms of GRs or GFs.
In such analyses, the active "subject" NP in the clause is typically jux
taposed with the nominal's pre-vN possessive NP, and the clausal "direct
object" NP is typically juxtaposed with the of-marked post-vN argument
NP. As has been demonstrated, not only is the double genitive assumption
unsound from a universal perspective (cf. 2.2.4), but it has negative ramifi
cations in analyses of the English vNP. Specifically in this last regard, the
discussion in 2.3.2.4 showed that the double genitive assumption diverts
attention from important semantic differences between nominals like the
investigation of Sherlock into the murder (unbounded activity) vs. the inves
tigation of the murder by Sherlock (delimited event). In regard to argument
preposing, the double genitive assumption creates another problem in
analyses hoping to define syntactic relations in the vNP by way of GRs or
GFs. It forces analysts to account in unnecessarily complicated terms for
the inability of Type A clausal-"subject" NPs to occur in the structurally
juxtaposed pre-vN possessive-NP position. With this large group of ACM
experiential-state nominals, which encode only the state created/induced by
the action or event denoted by the verb,^36 the clausal-"subject" NP cannot
occur prenominally:
(59) a. CL[The cowboyA:ef SUBJ amused SaraU:exp DOBJ]
b. * vNp[the cowboy's amusement of Sara]
Even in LFG, where not only is the clausal SUBJ GF explicitly distin
guished from the nominal POSS GF, but relating Ns to Vs by way of GFs
— even indirectly — is explicitly rejected, the implicit assumption inherited
from the double genitive assumption is that the argument linked to the
SUBJ GF in the clause will be linked to the POSS GF in the nominal. As
the example given in (59) demonstrates, Type A nominals refute the cor
rectness of that assumption.
Rappaport (1983) tried to resolve this problem by stipulating that V
sources of this nominal class have two argument structures — one with an
agent argument and one with an "experienced" argument (a thing which is
experienced), either of which may be linked to the SUBJ GF. Nominalized
counterparts, however, can inherit only the argument structure which
includes an experienced argument, and experienced arguments can never
be linked to the POSS GF. Although this ad hoc stipulation prevents the