ON DEVIANT CASE-MARKING IN LATIN^339tein (1976) and Maraldi (1983). Both analysts argue that these complements
represent single embedded sentential constituents. As mentioned earlier,
Maraldi regards them as equivalent in structure to such accusative-infinitive
clauses as (20b-c), likewise dominated by S'. Bolkestein argues (p. 285)
that since the accusative argument of the infinitival complement in question
"does not stand in any semantic relationship to the [matrix] verb, [it] con
sequently [does not stand] in a syntactic relation" to that verb. Bolkestein's
position is, I think, particularly indefensible; its guiding assumption is easily
falsified by such obvious counterexamples as English "raising to subject"
constructions, wherein the "surface subjects" of semantically single-place
predicates like be likely or seem require number and person agreement of
these verbs (The candidates is/are likely to be incoherent), despite lacking
a semantic relationship to them (as proven by the well known tests using
various semantically empty subjects, e.g., It is likely to rain as against It is
eager to rain). Bolkestein's assumption that the accusative argument of the
infinitival complements of verba sentiendi ac declarandi cannot bear any
syntactic relation to the matrix predicate prevents her from providing any
coherent account of the passive construction exemplified in (21a). She
argues (p. 282) that "information structure" may allow promotion to sub
ject position of an argument which does not "fulfill a syntactic function in
the main clause." This principle seems both vague and somewhat ad hoc.
Maraldi's account of such sentences as (21a) is also unsatisfying. The
case properties of both (21a) and (27) are explained in terms of S' erasure,
which allows the matrix predicate to govern the trace of NP movement, as
in (21a), or the subject of an embedded S, as in (27). This treatment forces
her to regard these two sentence types as manifestations of two different
constructions, both of which allow S' erasure to apply to their embedded
clauses. Such a treatment, however, obscures the following fact: among
verbs which take as their single semantic argument a proposition syntacti
cally represented by an accusative-infinitive complement, only those which
sanction the (core coordinate) structure in (27) permit the nominativus cum
infinitivo construction exemplified in (21a). Predicates like manifestum est,
which sanction core-subordinate accusative-infinitive complements (20c),
do not participate in the "raising to subject" construction: as shown above,
such sentences as (21b) are ungrammatical. It seems apparent, in fact, that
the nominativus cum infinitivo "construction" exemplified in (21a) simply
represents the passive version of that exemplified in (27), and that the
nominative-infinitive does not have the status of an independent grammati
cal construction in Latin.