ON DEVIANT CASE-MARKING IN LATIN^325Maraldi also applies the S' -deletion analysis to accusative-infinitive comple
ments of the type shown in (20a), it is evident that these embedded clauses
have a rather different status than do those in (20b-c). The accusative sub
jects of such embedded clauses, unlike those in (20b-c), appear to represent
core arguments of the matrix verba sentiendi ac declarandi (in (20a), dico).
When the matrix verb is passive, such arguments can serve as the subjects
of that matrix verb, appearing in the nominative and triggering agreement
of the passive matrix predicate. This situation is shown in (21a).
It appears that only accusative "subjects" of accusative-infinitive
object complements can become matrix subjects, and this only when the
matrix (transitive) verb is passive. As shown in (21b), the accusative "sub
ject" of the infinitival subject clause of manifestum est cannot serve as sub
ject of the matrix predicate. (As will be argued below, Latin does not sanc
tion a "raising to subject" construction.) The personal passive of (21a) can
be compared with an alternative impersonal passive construction
exemplified in (21c). In the latter case, the "subject" of the embedded
infinitival clause does not represent the subject of the passive matrix verb.
Instead, it seems that the entire accusative-infinitive clause represents the
subject of the agentless passive verb traditum est. Thus, the structure of
(21c) is analogous to that sanctioned by one-place active predicates requir
ing nonfinite sentential subjects (20b-c).
(21) a. Cicero consul esse ab eo dictus est.
Cicero(N) consul(N) be(iNF) by () said(N) is
"Cicero was said by him to be consul."
b. * Cicero manifestus est consul esse.
Cicero(N) evident(N) be(iNF) consul(N) to be
"Cicero is evidently consul."
Traditum est Homerum caecum fuisse.
said(N) is () blind(A) have-been(iNF)
"That Homer was blind is often said." Cic. Tuse. 5,114
With respect to foregrounding passives, Foley & Van Valin (1985:313)
argue that, although there do exist languages (e.g., Tagalog) with passives
in which peripheral (non-core) arguments serve as pragmatic pivots [PrPs],
such passives are exceedingly rare cross linguistically. The far more com
mon type of foregrounding passive is that in which only core arguments can
be "promoted" to PrP status. Thus, it seems that one can conclude from the
personal-passive form exemplified in (21a) that the accusative "subject" of