Advances in Role and Reference Grammar

344 LAURA Α. MICHAELIS

Regarding the first component of the passive statement, languages

vary as to which arguments they allow to be coded as pivots by the passive

linking, but universally, undergoer is the unmarked choice. (See Van Valin

1991, "Synopsis", sect. 5.5.) Latin, like English, allows the promotion to

PrP status of two types of macrorole-bearing arguments: an undergoer

included in a predicate's lexical representation, or an actor or undergoer of

an embedded clause which is linked to a syntactic position in the main

clause in a core juncture ("raising to object").

Regarding the second component of the passive rule, it seems that

some of the Latin data show that a broader characterization of the demo-

tional component of the passive rule is required for Latin. As shown in

(17c) and (17f), there exist impersonal passives in which an undergoer,

rather than an actor, is linked to the clausal periphery (recall that single

macrorole statives are held to assign the undergoer macrorole). Such a

characterization is suggested by Van Valin (class lectures), who argues that

voice alternations, which he holds to be fundamentally demotional, "in­

volve the non-canonical morphosyntactic coding of a macrorole, e.g.,

nonoccurence or oblique status." He further argues that languages vary

with respect to which macroroles they allow to be demoted. This variation

is expressed by an implicational hierarchy of the following sort:

Actors of transitive verbs > Actors of intransitive verbs >

Actors or Undergoers of intransitive verbs

English is held to represent the most restrictive type, while such languages

as Turkish, which allows passives of "unaccusative" verbs, are held to

exemplify the least restrictive type. Latin, it seems, lies somewhere in

between these types on the hierarchy. While one-place activity verbs form

impersonal passives (17a), one-place statives, e.g., nauseo ("I feel seasick")

do not appear to have passive manifestations; such impersonal passives as

*nausetur ("it is felt seasick") are unattested. Hence, it seems that Latin,

unlike Turkish, does not allow the demotion of undergoers of one-place

intransitives. As mentioned, however, two-place intransitives ("quirky

verbs" assigning only one macrorole) do appear to allow undergoer demo­

tion, as shown by the impersonal passives in (17c) and (f). Thus, it appears

that a relevant restriction upon macrorole demotion in Latin involves argu­

ment places: predicates having two or more places allow demotion of that

argument bearing the highest ranking macrorole, whether A or U; predi­

cates having one place allow demotion only of actors.