404 MARY L. NUNES
diction only with vNs derived from single argument activities (and plural
result vNs derived from iteratively-induced ACT senses of ACMs).
The activity nominals in the data base which do not appear to adhere
to the RRG prediction are two-argument vNs/vN-ing's. Not only are most
of these nominals unable to mark the CL-A as the direct argument, but
most of them are able to mark the y argument as the direct core argument.
In fact, attack is the only vN included in the data base which behaves in a
manner anticipated for activities:
(45) a. the attack [of the dogx:CL_A] on the childy
b. the attack by the dogx:CL_A on the childy
*the attack [of the childy] (by the dogx:CL_A)
d. the attack on the childy (by the dogx:CL_A).
Before assuming that the rest of the ACT vNs are exceptions to the
RRG prediction, it is essential to consider the behavior of investigation.
Notice how the interpretations of the macrorole assignments for Sgt. Brown
differ in (46a-c), and how the interpretation of the macrorole in (46d) is
ambiguous between A and U. (vN direct arguments are placed in brackets.)
(46) a. the (completed) investigation [of the murdery] by Sgt.
Brownx: CL-A
b. the (completed) investigation [of Sgt. Brownx.:≠ CL.A = υ]
c. the (* completed) investigation [of Sgt. Brownx:CL_A] into the
murder
d. ≠The investigations [of the subcommittee≠CL-A/CL-U] revealed
nothing about who had failed to report gratuitous income.
In (46a,b), the direct argument of investigation is interpretable only as a
non-actor — i.e. as the entity undergoing, rather than performing the inves
tigation. In these constructions, the adjective completed, which can only be
used to describe bounded events including a telic end point which has been
reached, may modify investigation. Thus, (46a,b) behave in the manner
predicted for ACM-based vNPs.
In (46c), however, completed may not be used to describe investiga
tion. The additional context of into the murder prevents the state of affairs
named by the nominal from being interpreted as a bounded event with a
telic end point which has been reached. Furthermore, the direct argument
of the vN can only be interpreted as an A — i.e. as the entity performing,
rather than undergoing the investigation. In this case, investigation behaves
in the manner predicted for activity vNs.