Advances in Role and Reference Grammar

ARGUMENT LINKING IN DERIVED NOMINALS 403

Test 4: Jeri finished analyzing the data.

OK if a telic end point — determining the nature of the data as

a whole — was reached, such that there was nothing left to

methodically examine for the purpose of reaching that end:

= ACM

BAD if no telic goal was realized, such that the absence of a

culmination point (defined in terms of the nature of the data

as a whole having been determined), precludes the possibil­

ity of there being nothing left to methodically examine: =

ACT

Thus, when the atelic sense of methodical examination is assumed, analyze
tests as an ACT. However, when the telic sense of determining the nature
of the whole is assumed, the boundedness of the event alters the tests'
results to that of an ACM. Without either context, Jeri analyzed the data
cannot be effectively tested. Some of the other verb sources included in the
ACT data (e.g. investigate) pattern similarly to analyze when these tests are
applied.
The ability/inability of a clause to passivize is another indicator of a
verb's ACT vs. ACM class-sense status. Whereas constructions headed by
ACM verbs passivize readily, those headed by an ACT are difficult to pas­
sivize, except in constructions with a generic Italians eat spaghetti; Spaghetti
is eaten by Italians). Thus, passivization is actor (e.g. permitted in the
clauses below only where the predicates refer to definite caused-and-
bounded events, as in (44)).
(43) ACT: a. Spaghetti was eaten by Joe for five minutes.
b. Places were explored by Ponce de Leon in the 1500's.
 * Decisions were criticized by Sue.
(44) ACM: a. A plate of spaghetti was eaten by Joe.
b. Florida was explored by Ponce de Leon in the 1500's.
 The Supreme Court's decisons were criticized by Sue.
Crucially, the susceptibility of ACT verbs to ACM interpretations
appears to have ramifications in the vNP. In turning to these ramifications,
it is important to recall that RRG predicts that ACM-based vNs will select
the argument linked to the CL-U as the direct argument, while ACT-based
vNs, having access only to the intransitive verbs' single macrorole for such
linking, will select the argument linked to the CL-A as the direct argument.
As has already been noted, the accomplishment data straightforwardly ver­
ify the RRG prediction, but the activity data transparently verify the pre-