368 LAURA Α. MICHAELIS
itive two-place and transitive three-place predicates. Jensen's assumption
that the "quirky" two-place predicates deviate with respect to their case per
se prevents the recognition that their case patterns manifest general princi
ples governing the coding of both second and third arguments.
An additional respect in which Jensen's analysis seems somewhat
unsatisfactory is in its treatment of such variable valence verbs as dono (2a,
14c). As discussed, such verbs are regarded here as licensing two linkages
of the thematic role theme to the macrorole of undergoer, one marked and
one unmarked. In the case pattern exemplified in (2a), the theme is linked
to U and hence receives accusative case by the coding principle (24b); the
locative accordingly receives dative by (24c). The case pattern given in
(14c) exemplifies the marked linkage of (26), in which the theme argument
receives ablative owing to its being outranked for U by a locative argument
which is thus realized as accusative by (24b). Variable-valence verbs in
Latin are hence explained as those licensing the marked linkage of (26) in
addition to that linkage predicted by the A/U hierarchy. Jensen's account
of variable valence, however, relies upon the rather dubious claim, last
embraced in transformational accounts such as Ross (1967), that in Latin
"the order of the noun phrases following the verb is important for the syn
tactic representation," and that "the free word order generally ascribed to
Latin is the result of a relative freedom of scrambling on the...branch lead
ing from syntax to the phonetic form" (p. 15). To explain the dual sub-
categorization of dono, Jensen must claim that in the lexical entry for this
predicate, only the first noun phrase following the verb can receive accusa
tive case via government. This noun phrase may bear the thematic role fea
ture [+goal] (locative) or [—goal] (theme); the oblique argument will then
bear ablative or dative case in accordance with its thematic role (ablative if
[-goal], dative if [+goal]). As tidy as this analysis may be, the claim that
the presence of accusative case here is somehow a feature of word order
imposes upon Latin constraints for which evidence is lacking. Since the pre
sent analysis of variable valence does not impose upon lexical entries in
Latin such poorly substantiated ordering constraints, it appears a more
plausible account of this phenomenon.