496 MICHAEL SILVERSTEIN
tive case-marking for a relatively lower "A", as in stratum I; the rest of the
paradigm is then generated, as in stratum II, by pre-posing "A" or "S" that
remain (obviously a kind of topicalization) in two forms, the special erga-
tive form only in the nonsingulars. The languages really differ, we are
claiming, in two ways: First, the Djirbal normal forms are "direct", they
code the "A/S" case-relations with nominative/absolutive and the "O" case-
relation with dative; the Chinookan normal forms are "inverse", they code
the "A/S" case-relations with dative and the "O" case-relation with nomina
tive/absolutive. Second, Djirbal has all three case-relations available as
maximally tight coreferents (coded with zero); Chinookan has only "A"
and "S" available for this type of linkage.
Contrast a language like English, as a (somewhat odd, to be sure, but
usable) representative of a fairly thoroughgoing nominative : accusative
basic system. To be sure, we do not have an antipassive in English, but
rather a passive, which is used in a number of contexts where the other lan
guages use antipassive, namely, to allow maximally tight discourse corefer
ence linkage. The passive, of course, expresses the "O" noun phrase in
nominative/absolutive form, and optionally expresses the "A" noun phrase
in some adverbial or lexically dative form (dative 2 ). As far as I can see, it is
at the level of "desire complement", e.g., The man wants s( ) and above on
our hierarchy of linkage that zero reference maintenance becomes possible
in English with use of the nominative/absolutive: dative schema, that is, the
passive, which obligatorily is used to make coreference with the Patient (or
"O") case-relation in the linked clause. Thus English can be brought into
the typological comparison as well, with the normal form of inflection hav
ing nominative/absolutive : dative 2 case-marking where nominative/absolu
tive represents "S" and "O" and dative represents "A". All three of these
case-relations are available for maximally tight coreference (with zero) and
the intervening D 1 as well. For plain forms of inflection in stratum II, the
Patientive hierarchy applies to all "O" forms and marks them with a special
accusative (or the same as the dative), and some "A" forms, as function of
verb and lexical hierarchy, are topicalized into nominative/absolutive. Eng
lish, unfortunately, does most of its equivalent transformational work with
lexicalizations of alternative case-relation codings, rather than with sys
tematic morphosyntax. But where it shows the relevant systematic distinc
tions, they are entirely consistent with what we have already outlined.
Recall, the argument has been that there are central case-markings
both that surface in some particular, predictable syntactic environment, and