OF NOMINATIVES AND DATIVES 475cates and arguments. Another of these domains is clause-clause logical rela
tions, which involves a higher level of representation not clearly of the same
type (unless you uninterestingly call such relations as "temporal sequential-
ity" higher predicates, solving the problem by notational fiat), and expres
sed differently in various languages with compound and complex sentences,
paragraph and discourse structures, etc. The last of these domains is refer
ence-maintenance, the introduction of identifiable entities as ultimate argu
ments of discourse, and the continued separate conceptualization of them
as entities, marked in the overt structure of discourse; such phenomena as
so-called "topic" are of this domain. (Further interactions of so-called case
and other categories are also documented, but I do not take them up here.)
Case-marking, as I said, is at the intersection of these various domains,
in different, but consistent ways in every language. That is, case-marking is
not a functionally pure phenomenon from which, even with a bit of trans
formational hand-waving, we can read anything about underlying coding
constraints. We must examine the interaction of all of these domains in any
account of the facts of distribution and interpretation. Of course, that case-
marking, unlike phonological inventories and distributions, is not a
unifunctional phenomenon, i.e., can not be significantly brought into order
with a single functional hypothesis, such as the inadequate and ill-fated sub
ject-object schema of certain transformationalists, makes it more compli
cated to deal with than phonology. But that is just a fact of life about actual
languages as mechanisms for coding communicable content, the signifi
cance of which we could, but will refrain from, speculating about.
Let me characterize the kinds of dependencies that are well-established
between each of the four functional domains and surface case-marking. Let
us take them in the following order: inherent lexical content, propositional-
ity, then two that are linked (for obvious reasons), clause-linkage and refer
ence-maintenance. Where I single out one of these domains, it is to be
understood that all other things remain equal, under the generalization.
Noun phrases are differentiated by a hierarchy of features (or equiva
lent) that intersect to define a hierarchy of noun-phrase types or categories
to which the case-marking schema of any language is sensitive, in unencum
bered "kernel" inflection. The feature hierarchy is given in Figure 4. This as
I term it "plain" inflectional schema for propositional Agents and Patients
and Subjects (only the last, single-argument propositional form is so
termed) works according to two principles, called the Agentive inflectional
principle and the Patientive inflectional principle, each of which can be sen-