384 MARY L. NUNES
2.2.2 The inherent "intransitivity" of nominals
In the clause, there is typically a correlation between the number of mac-
roroles taken by a verb and the number of direct core argument slots. That
is, if a verb takes two macroroles, it usually has two direct core arguments;
if it takes one macrorole, it has one direct core argument. Exceptions to this
in English occur in 1) marked dative shift constructions in which the verb
has three direct arguments, only two of which are linked to macroroles (cf.
2.3.1 and 2.3.3 for discussion), and 2) constructions headed by an activity
verb with two direct arguments, only one of which has macrorole status.
In nominals, however, there is never more than one direct core argu
ment, regardless of how many macroroles there might be. Thus, nominals
are inherently intransitive in the traditional syntactic sense of transitivity.^9
This is true even in vNP constructions where all of the arguments included
in the semantic representation are present. Only one argument may be
linked via a macrorole to the single direct core argument position following
the nuclear vN and marked by of. vNs do not randomly select an argument
for the direct core argument position. Rather, the selection is predictable in
terms of the following linking hierarchy:
(8) Direct-Core-Argument Linking Hierarchy: U > A
The hierarchy predicts that with vNs related to state [STA], achieve
ment [ACH], or accomplishment [ACM] verbs taking two macroroles, the
argument which may be linked in the vNP to the direct core argument posi
tion is the argument associated with the U. With vNs related to verbs taking
only one macrorole, the argument which may be linked in the vNP to the
direct argument position is the argument associated with that macrorole.
This, of course, predicts that the direct core argument of vNs related to
activity [ACT] verbs, which take only As, will be the argument associated
with the A. It is particularly noteworthy that in the RRG framework, the
macrorole markedness hierarchy of U > A for the privileged syntactic posi
tion in the clause characterizes ergative languages, whereas A > U charac
terizes accusative languages.^10 In the English clause, the linking hierarchy
for the privileged syntactic position works accusatively. However, as
Sadock and Levi (1977) and Levi (1978) have pointed out, and, as the link
ing hierarchy given in (8) reflects, the English nominal in general exhibits
ergative patterning. That is, the linking hierarchy for the NP's privileged
syntactic position (the direct core argument position) works ergatively with