ON DEVIANT CASE-MARKING IN LATIN^331text, would be coded by a nominative argument. Sentence (25d)
demonstrates both that the nominative NP (here, Metellus) controls zero
anaphora in a paratactic structure, and that the missing arguments of the
conjoined predicates {donat, laudai, and agit ) are reconstructible only as
nominative NPs. Hence, the nominative NP (most typically) represents
both the trigger and target of zero anaphora.
We have thus seen a number of constructions to which the grammatical
function represented by the nominative argument is central. The nomina
tive argument, we have seen, is "raised to object", typically controls the
reflexive adjective sum, is the "target of deletion" in "equi" structures,
and, where zero anaphora occurs, ordinarily provides both the antecedent
and the reconstruction of the null anaphor. It should be noted here that the
use of the terms "raising" and "equi" obscures a crucial similarity between
the two types of structures; in both, a missing complement of an embedded
infinitival predicate receives its interpretation through coinstantiation with
an argument fulfilling a valence requirement of the matrix verb. (See
"Synopsis", sect. 7.2.) In both cases this coinstantiated argument represents
the grammatical function coded by the nominative in main clauses. The
nominative then appears to serve a pivotal grammatical function with
respect to much of the syntax of Latin.
What remains to be demonstrated, however, is that this pivot is one
whose selection is determined by discourse-pragmatic factors, particularly
topichood. Such factors must be separated from those related to a seman
tic-role hierarchy. In Latin, as in English and German e.g., that NP bearing
the actor macrorole will be the preferred choice for pivot. Thus, as shown
in (20a) and (25) it is typically an actor which "raises to object", controls
the reflexive possessive adjective suus, provides the "target" for "subject-"
and "object-control equi", and serves as both antecedent and null anaphor
in constructions allowing zero anaphora. Latin, however, differs from such
languages as Enga and Walpiri (Foley & Van Valin 1985) in the following
respect: the actor need not always serve as pivot. In Latin, an alternative
choice of pivot — that of undergoer — can be made, by dint of the passive
construction. And, as argued by Foley & Van Valin (1985:305), "voice dis
tinctions of the type active-passive are a common feature of languages
allowing undergoers to serve as pivots." In Latin, as in English (cf. van
Oosten 1984), the placement of an undergoer in pivot position seems to
provide for discourse cohesion; it allows one to reserve subject position for
a particular sentence topic across clauses, whether it functions as actor or