Advances in Role and Reference Grammar

ATTRIBUTIVES AND IDENTIFICATIONALS 435

tion will be semantically motivated (in the sense of Lakoff (1987)). This is
the position which I will adopt for this paper. Particularly interesting to
proponents of semantic alignment hypotheses is the case where an identifi­
able semantic class of predicates appears to systematically fall into the
wrong syntactic class. An apparent instance of this sort of situation is men­
tioned in Rosen (1984), where Italian verbs of motion with a specific goal
pattern like verbs of the objective class in taking the auxiliary essere "to be"
rather than avere "to have", despite the fact that they are understood to
involve protagonist control and despite the fact that in other languages,
such as Albanian and Choctaw, they pattern like verbs of the subjective
class. Thus, for example, in (la), corso takes the auxiliary ha {avere) when
it occurs without a specific goal, but in (lb), it takes the auxiliary è {essere)
when it occurs with a locative goal.

(1) a. Ugo ha corso meglio ieri.

Ugo has run better yesterday

"Hugo ran better yesterday."

b. Ugo è corso a casa.

Ugo is run DIR home

"Hugo ran home."

However, Centineo (1986) and Van Valin (1987, 1990a) argue that in Ita­

lian, it is the relation of the subject to a stative predicate in the Logical

Structure of the clause which determines the syntactic class of these verbs,

rather than protagonist control. Motion-to-goal verbs are analyzed as

accomplishment verbs which have a stative predicate in their Logical Struc­

tures indicating the final state of accomplishment, so that motion-to-goal

verbs have logical representations containing the stative predicate be-at'.

Motion verbs without specific goals have logical representations which do

not contain a stative predicate. This analysis will be examined further in

section 3. Van Valin (1987, 1990) further argues that subjectivity and

objectivity cross-linguistically are determined by language-specific paramet­

ers of verbal semantics. This hypothesis will accommodate, and in fact pre­

dict, cross-linguistic variation in the membership of the objective and sub­

jective syntactic classes, as long as that variation is statable in terms of the

lexical semantics of verbs.

Another instance where a semantic class of predicates does not seem to

fall neatly into the previously-identified semantic parameters determining

the subjective-objective split in languages, and the one which is the topic of