486 MICHAEL SILVERSTEIN
ous telescopings and elaborations of the case-marking system of any lan
guage form a regular array from least elaborated to most elaborated, in
terms of the four domains of variables in their formal aspects. Further, that
the nominative/absolutive : dative opposition is, in a sense, the fundamen
tal opposition of languages in their grammatical case-relation systems, from
which are elaborated other grammatical case-markings. Further, that to the
degree to which peripheral, adverbial cases can be regularly related to
grammatical ones, they will be related to the dative as a minimum, and to
the nominative/absolutive secondarily. Further, that it is not some abstract
underlying case-relation called "nominative/absolutive" and "dative" that
we are talking about, but the very ones that surface in the least linked, max
imally diverse plain inflectional structures that also have this foundational,
anchoring effect in terms of the whole structure. They are the foci or focal
points of the case-marking system. Thus, the actual case markings are sub
jected to a strictly empirically recoverable formal-functional hypothesis,
unlike theories of case grammar, which float free in some semantic never-
never-land, without surface constraint, or theories of relational grammar
and other purely formal notions of syntax, which have no restrictive
hypotheses about the possible interrelations of actual surface shapes, not of
some postulated but abstract zero representations.
In Djirbal, for example, there is a combination of several interesting
factors from this point of view, in which the nominative/absolutive : dative
opposition shows these typological characteristics. Djirbal is a case-marking
language in the classic sense, in that noun phrases have morphologically
expressed case desinences in much the same way as Sanskrit or Estonian.
Lexical nouns have absolutive, ergative, dative, genitive grammatical case,
and locative, allative, and ablative adverbial cases, while pronouns of the
first and second persons have nominative, accusative, dative, genitive
grammatical cases. The lexically-governed split of ergative : absolutive vs.
nominative : accusative case marking systems is neatly at the boundary
between the true pronouns (first and second person) and the third person
noun phrases. In the plain inflection of independent "kernel" clause-sen
tences, the inflection is according to whatever the lexical content of the
respective noun phrases requires. In addition to this form of clause, how
ever, there is, as in most somewhat ergative systems (note that ergativity vs.
accusativity is a matter of degree, not of all-or-nothing choice), an antipas
sive form of clause, in which the mirror-image formal transformation from
the passive of accusative languages occurs (see Figure 9). The Agent case-