474 MICHAEL SILVERSTEIN
2. Case-marking categories in particular
Let us turn to another formal-functional domain, which was originally of
particular interest to me for a number of reasons. The phenomena of case-
marking, the indication of underlying propositional case-relations of predi
cates and their arguments, however indirectly but systematically, are fas
cinating precisely because, especially for the traditional so-called "gram
matical" cases, they seemed to be the most abstract, formal, syntactic (as
opposed to semantic) transformational markers, on the basis of which one
could really understand the abstractness and autonomy of morphosyntax.
Thus, "subject" and "object," as the two central characterizations of syn
tax, were, and unfortunately still are in some circles, the foundation of so-
called thematic relations preserved in transformational shifts, etc. The exis
tence of what we call "ergative" case-marking systems, with a number of
mirror-image syntactic properties from the perspective of the usual "accusa
tive" systems, as we have come to be able to characterize subject-object
orientation, thus presents an analytic challenge, the exploration and expla
nation of which in an orderly fashion ultimately yields the same formal-
functional insight as any other domain of grammar. That is to say, the only
way adequately to order the phenomena of case-marking is by formal-func
tional universals that deny the autonomy of these, once-seeming maximally
abstract forms in some separate syntactic component and to see the coding
properties of distinct case-markings as at the intersection of a number of dis
tinct semantic-pragmatic domains, among them the four given in (1).
(1) Case Marking:
Independent Variables: Dependent Variables
a. Inherent lexical content of NPs (see Figure 4) J
b. Clause-level propositionality (see (2)) f Type of case-marking
Clause-clause logical relations (see Figure 6) Í oppositions
d. Discourse reference-maintenance (see Figure 7)1
One such domain is the domain of local inherent lexical content, however
we might want to try to represent it (e.g., with features, one-place or more,
with definitional equivalences, with various extensional metalanguages
where this is possible, etc.), where noun phrases in overt structure seem to
differentially and cross-categorially divide up an entire universe of possible
entities for reference. Another of these domains is clause-level proposition
ality, the interpretability of explicit clause structure (and what can be moti
vated as equivalent by transformational means) as configurations of predi-