338 LAURA Α. MICHAELIS
Such sentences as (21c) are not, however, problematic for the clause-
linkage analysis. A tenet of this theory is that a given syntactic structure can
reflect one or more clause-linkage types. Thus, one need not claim that
core coordination is the only clause-linkage type instantiated by accusative-
infinitive complements; instead, the infinitival clausal subjects of such one
place predicates as constat and manifestum est (20b-c) appear to exemplify
core subordination: a unitary S-constituent serves as the core argument of a
predicate, with no arguments within it bearing any relation to that matrix
verb. It is apparent that verba sentiendi ac declarandi sanction two passive
constructions, a personal passive instantiating core coordination and an
impersonal passive, structurally analogous to the active sentences in (20b-c)
instantiating core subordination. Only in the former case need we speak of
a "shared" core argument.
As shown earlier, this shared argument must be the highest macrorole-
bearing core argument in its clause; were the accusative-infinitive object
complement a finite sentence, the shared argument would be PrP, and
hence nominative. As an argument of the matrix verb demiror, however, it
is a non-PrP macrorole-bearing core argument, and it hence receives
accusative case in accordance with (24b). Its receiving accusative case does
not depend on its occupying a particular macrorole; in (27), the "raised"
accusative argument of the higher verb is an actor in its clause; in (23f), the
"raised" accusative argument is an undergoer in its clause. Furthermore,
the assignment of accusative case to this argument is not dependent upon its
serving as grammatical object of the matrix verb. The case-marking princi
ples in (24) are not sensitive to any grammatical function other than prag
matic pivot; (24b) simply assigns accusative case to a non-pivot macrorole
core argument, whether that argument has this status by dint of core coor
dination or lexical subcategorization. Although (24b) does not directly
account for the assignment of accusative case to the "subjects" of core sub
ordinate infinitival clauses like (20b-c), we might adopt some version of the
case-marking principle suggested by Bolkestein (1979:32), "the accusative
case form is assigned to...subjects [of non-finite verbs] instead of the
nominative case form which is assigned to subjects of finite verbs forms." It
may be that, in RRG terms, Latin infinitival sentential subjects do not
assign PrPs, in which case (24b) provides a ready account of the case of
their "subjects".
The clause linkage analysis of the accusative-infinitive complement-
type exemplified in (27) can profitably be compared with those of Bolkes-