ON DEVIANT CASE-MARKING IN LATIN 343accusative case-forms with the predicates involved, we also find "normal"
cases with the accusative, both before and after the classical period...In
fact, in the course of history the accusative prevails with most of the verbs
[sanctioning non-accusative case-forms]." Further, it appears that even
within the classical period, "quirky" verbs like noceo have occasional trans
itive manifestations; Pinkster (ibid) cites such instances as Caesar Civ. 1,
- Hence, Latin appears to manifest both a synchronic and diachronic ten
dency toward elimination of deviant case through leveling, a tendency
which suggests that deviant case-patterns are indeed idiosyncratic.
Thus, it appears that although most "quirky" verb classes examined
here are atypical with respect to Rice's prototype, it does not appear that
there is any one-to-one relationship between a predicate's case pattern and
its degree of prototypicality with respect to the transitive-event scenario.
Rather, it appears that deviant case-marking need not reflect profound
deviance from a semantic transitive prototype, nor must such deviance be
manifested as quirky case. The intransitivity feature held here to underlie
deviant case might thus be treated as a syntactic quirk of individual predi
cates, one which must be noted in their lexical entries. We must then return
to the question of how we might represent this sort of idiosyncracy in the
lexical entries assigned these verbs.
In RRG, transitive verbs are defined as just those predicates licensing
the two macroroles actor and undergoer. In the default situation, as stated,
verbs licensing two or more direct core arguments license these two mac
roroles. Deviant case-marking verbs assign one less macrorole than one
would expect, given their number of arguments. The fact that a two-place
verb licenses a non-accusative "object" or non-nominative "subject"
reflects the fact that the particular predicate assigns no macrorole to this
argument.
The ability of a particular predicate to form a personal passive is a well
established acid test for transitivity. The claim that two-place verbs sanc
tioning non-accusative objects can be characterized as intransitive is sup
ported by the fact that such verbs form only the impersonal passives
exemplified in (17). Passive formation in RRG involves promotion of the
undergoer to subject position, with concomitant "demotion" or
backgrounding of the actor. The RRG formulation of passive represents
not a lexical rule, but a linking algorithm specifying the manner in which
macroroles are mapped into grammatical functions. It is given the following
(universal) formulation in Van Valin (1991): ~ A [universal default=U] =
Pivot, A= X (A is omitted or linked to a peripheral status).