260 WILLIAM H. JACOBSEN, JR.
retained only on the first verb of a sequence, and (Mayan) Jacaltec, where
the non-past tense marker is understood to convey the same tense that is
marked on the following verb form.
The introduction of this third possibility alongside of traditional coor
dination and subordination seems useful at a certain level of abstraction.^21
Yet cosubordination is not sharply opposed to them, as our authors point
out with respect to the properties of embedding and dependency. Cosubor
dination is thought of as being intermediate between coordination and sub
ordination, sharing with the former the feature [- embedded] and with the
latter the feature [4-dependent] (Foley & Van Valin 1984:241-243; Van
Valin 1984:546-547).
I find a lack of symmetry here, however, and have come to feel that
(clausal) cosubordination is best regarded as coordination with the added
property of operator dependency, perhaps better called operator projec
tion, whereby the clausal operators (such as tense, mode, evidentiality) of
one clause are understood as applying also to certain nearby clauses whose
verbs carry no contradictory indications.^22 This is all the more so for the
Role and Reference Grammar framework, which does not require a con
junction as a diagnostic characteristic of coordination, not therefore making
a primary distinction between conjunction and parataxis (Foley & Van
Valin 1984:243-244). The suggested equal closeness of cosubordination to
subordination is based on dependency, which falls into two types, distribu
tional and with respect to grammatical categories (Van Valin 1984:544-
547). The distributional limitations of cosubordination, not involving
embedding, are at most slightly more than those of coordination, including
linear directionality between the clause bearing the specifications of clausal
operators and that onto which these are projected. With regard to depen
dency of grammatical categories, I think a sharp distinction must be made
between, on the one hand, their projection onto another clause, as seen in
cosubordination and, on the other, their neutralization because the asser
tion is presupposed, as seen in subordination (a distinction indeed recog
nized by Foley & Van Valin 1984:240, Van Valin 1984:548).^23
One should also be alert to possible additional interactions among
these nexus types. Coordination of subordinate clauses is, of course, tradi
tionally allowed for. Coordinated subordinate clauses formed with ?ani
(section 10.7), without any conjunction, seem to occur in Nootka (28:6-7,
132:29-30). Other instances of coordination may group together cosubordi-
nated nuclei (21:1) and perhaps absolutive prepositional cores (132:1-2)
(again unmarked by any conjunction).