356 LAURA Α. MICHAELIS
ment principles in (24), one can conclude that such two-place predicates as
pudet lack the expected two macroroles. Macrorole-bearing arguments
must appear in either the nominative or accusative case, and only the
accusative is represented in the case frames of such predicates, coding the
cognizer. That the single macrorole-bearing argument appears in the
accusative rather than in the nominative indicates that these verbs license
no PrP. The claim that the genitive argument is not a PrP is corroborated
by the fact that it does not trigger verb agreement; all of the verb forms
licensing the genitive subject are impersonal.
Despite the fact that such verbs as pudet license no PrP, one might wish
to give them a treatment similar to that accorded the inverse verbs of (3),
which likewise sanction the linking of the cognizer role to a non-subject
position. If verbs such as pudet were analyzed as statives akin to placet and
libet, the case of the genitive "subject" would be explicable according to
principle (26): the single undergoer macrorole assigned by such predicates
would be occupied by the locative (cognizer) argument instead of the theme
(percept), and the latter would accordingly receive genitive case.
Such a treatment would not, however, allow one to explain why, as
shown in (23), the genitive "subject" appears to exhibit PrP-like properties
with respect to the accusative-infinitive object complement ("raising to
object" construction). It participates in this construction in the manner of a
nominative subject, although it retains its deviant case when "raised",
rather than appearing in the accusative. Hence, such sentences as (23a,c,e)
present a problem: if the claim is to be that all "subjects" or "objects" car
rying deviant case have not been assigned macrorole status (i.e., are not
linked to either actor or undergoer), and if macrorole status is to be an
absolute precondition for PrP status, then one has no means by which to
account for the fact that the genitive arguments in these examples display a
behavioral property apparently associated only with PrPs. That is, if the
genitive arguments in (19) and (23) are not to be considered actors, and if
non-macrorole core arguments are precluded from becoming PrPs, how can
we explain the fact that, in (23), they seem to have achieved, to some
degree, PrP status?
Within RRG, there is a means by which to account for those items
which "act like" subjects but which otherwise lack salient subject proper
ties. As mentioned above, in addition to a PrP, which is selected by dis
course or pragmatic factors and to which traditionally recognized subject
properties accrue (e.g., nominative case-marking in nominative-accusative