ON DEVIANT CASE-MARKING IN LATIN 337
algorithm account for the properties of the "normal" accusative-infinitive
complements exemplified in (23)? Let us take sentence (23b) as an exam
ple, repeated here as (27):
(27) Eum Gaium non monuisse ut iusseram
() Gaius(A) not to-have-warned(INF) as (I)ordered
demiror.
(I)cannot believe.
"I cannot believe Gaius not to have warned him as I ordered."
As was argued above, in the accusative-infinitive complement construction
exemplified in (27), the highest ranking macrorole-bearing argument in the
lower clause represents a direct core argument of the higher clause, as evi
denced by the fact that the accusative "subject" of the embedded infinitival
clause can appear as the nominative subject of the passive matrix verb. (Re
call that the accusative-infinitive complements exemplified in (20b-c) do not
have the property of sharing a core argument with the matrix verb; as
shown in (21b), their accusative "subjects" cannot serve as matrix subjects.)
Within RRG, the construction underlying (27) cannot be analyzed as
an instance of "raising"; underlying levels of syntactic representation are
not a feature of this framework. Instead, this construction is viewed as
exhibiting a type of clause linkage whereby the matrix and embedded
clauses share a core argument. (See "Synopsis", sect. 7.2.2.) "Raising to
object" can then be described as a situation in which the highest-ranking
macrorole-bearing core argument in an embedded clause serves as a non-
pivot macrorole-bearing argument in the matrix clause. Hence, within
RRG, the account of clause linkage needs neither movement rules nor mul
tiple levels of syntactic representation to account for the structural proper
ties of the accusative-infinitive construction shown in (27).
Aside from being somewhat more elegant, the clause linkage analysis
has an additional advantage over the raising-to-object account. As pointed
out by Bolkestein (1979), passive sentences of the type shown in (21c) are
problematic for the raising account. In contrast to (21a), (21c) has the
entire clause as subject of the passive matrix verb. It then appears that this
clause, rather than its accusative subject, represents a core argument of the
matrix verb. The raising-to-object analysis would predict that the type of
passive construction shown in (21c) does not exist, and in fact, Pepicello
(1977), in arguing for the raising analysis, simply ignores the extremely
common sentence-type exemplified by (21c).
