340 LAURA Α. MICHAELIS
That Latin has no "raising to subject" was first observed by R. Lakoff
(1968) with respect to adjectives, and appears to be true as well for one-
place verbs like constat ("it is agreed"); these verbs have only the imper
sonal uses exemplified in (20b-c). Bolkestein (1979) mentions such appar
ent instances of subject raising as ea oportent fieri (these things (N) ought
(3-PL) to be done), in which the one-place predicate oportet, an impersonal
verb having the valence properties of constat, appears to manifest the per
son and number agreement symptomatic of subject raising. Such cases, I
think, represent instead an alternate personal valence of the impersonal
oportet, perhaps originating via analogy with the two-place predicate debeo
("I must"). It then appears that the nominative-infinitive "construction" is
not represented in the grammar of Latin other than as the passive manifes
tation of the accusative-infinitive construction shown in (27). Neither
Maraldi's analysis nor Bolkestein's can account for the fact that the
nominative-infinitive "construction" of (21a) apparently owes its existence
to the accusative-infinitive construction of (27).
RRG, however, provides a straightforward account of the relationship
between the nominative-infinitive "construction" shown in (21a) and the
accusative-infinitive construction exemplified in (27). Because predicates
like dico sanction core coordination, the accusative argument Gaium of
(27) represents a macrorole-bearing core argument of both monuisse and
dico. This argument can hence serve as PrP in the matrix clause by dint of
the passive linkage whose universal formulation will be given below. Thus,
the RRG account of such accusative-infinitive complements as (27), unlike
those of Bolkestein and Maraldi, also provides a relatively coherent
account of the passive sentence-type exemplified in (21a).
Having looked at the RRG case-making algorithm with respect to the
those verbs licensing "normal" case-marking patterns in both complex and
simple clauses, we might now examine its application to those verbs which
license deviant case-patterns, but which do assign a PrP. (This latter condi
tion is designed to exclude the genitive-subject cases of (19) and (23) from
our immediate purview.) Before turning to the representations of these
verbs, we must explore a claim central to this analysis: that two-place pred
icates sanctioning irregularly coded "subjects" and "objects" are not irregu
lar with respect to their case-patterns per se, but with respect to their trans
itivity, a property which, according to Van Valin (1991), represents "an
area of notorious lexical idiosyncracy."