ON DEVIANT CASE-MARKING IN LATIN^345Having investigated the manner in which macrorole demotion is
restricted in Latin, we might now turn to an examination of how the two
components of the passive rule account for the data in (17) and (18). The
passive algorithm accounts for such personal passives as (18b) in a
straightforward manner: the argument linked to the actor macrorole, Cic
ero, is relegated to peripheral status, and is hence coded by an a-phrase;
the argument bearing the undergoer macrorole, Pompeium, is linked to
PrP, and hence both triggers verbs agreement and receives nominative cod
ing. Non-accusative objects, however, as shown in (17), cannot assume sub
ject status when the passive linkage is in effect; they neither trigger agree
ment nor receive nominative coding. This fact can be shown to lend cre
dence to the theory that such objects, although core arguments of their
predicates, do not bear the undergoer macrorole.
As shown, the RRG linking rule for passive involves both a foreg
rounding and a backgrounding component. The verbs forming the imper
sonal passives in (17) can be said to license only that component whereby
the subject is relegated to the clausal periphery (as shown in (17d), imper
sonal passives permit ab-phrases coding demoted actors). The foreground
ing component, whereby the argument occupying the undergoer macrorole
assumes PrP status, is inoperative here, for the following reason: verbs of
this class do not assign the undergoer macrorole to their non-subject argu
ments, and macrorole status is an absolute precondition for PrP status. The
non-accusative non-subject argument thus fails to receive nominative case
and instead retains its quirky case when the verb is passive.
One might note that this conception of the impersonal passive also pro
vides a straightforward account of the impersonal passives licensed by such
one-place activity predicates as curro, shown in (17a). These impersonal
passives are identical to those licensed by such "quirky" predicates as noceo
("I harm"), except, of course, for the presence of an additional core argu
ment. Such impersonal passives as (17a), like (17d), involve the
backgrounding of an argument bearing the actor macrorole, with no con
comitant foregrounding of an undergoer, which in (17a), as in (17d), is
missing. Hence, the fact that two-place "quirky" verbs and one-place pred
icates have identical passive manifestations can readily be accounted for
within the present framework.
It might be prudent at this point to defend the intransitivity analysis of
"deviant" verbs against charges of circularity recently levied against it by
Pinkster (to appear). He gives the following argument against this analysis: