Advances in Role and Reference Grammar

ON DEVIANT CASE-MARKING IN LATIN^361

accusative-infinitive construction admits of both pivot types, a PrP, when
present, will be given preferential treatment.

3. Comparison with previous analyses

The advantages of the present analysis can best be seen by comparing it
with several earlier attempts to account for deviant case-marking in Latin.
In this section, three accounts of this phenomenon — authored by Pinkster
(1985, to appear) and Jensen (1983) — will be summarized and compared
with that proposed here. I hope to demonstrate that, while each of these
analyses offers some intriguing insights into the phenomena under investi­
gation, the present analysis is to be preferred both for the range of data it
accounts for and the applicability of the coding principles suggested to the
Latin case system in general.

3.1 Pinkster (1985)

Both this analysis and the present analysis represent attempts to provide a

unified account of the case-marking of "third arguments" and non-accusa­

tive "second arguments" (i.e., the dative coding of both indirect objects

and most non-accusative objects, as well as case-pattern parallelisms

between certain statives sanctioning ablative and genitive objects and their

causative counterparts). Both analyses recognize that certain semantic sub­

classes of verbs (of lacking, etc.) are associated with a case-pattern charac­

terizing both "quirky" two-place Stative verbs and three-place causative

verbs.

There is, however, a fundamental difference between the two

accounts: while the latter regards deviant case as an idiomatic feature of the

verbs sanctioning it, the former regards it as synchronically motivated. In

Pinkster's analysis, two "semantic justifications" for the presence of dative

and ablative non-subject arguments are adduced. Pinkster first advances

the following "quick generalization" for three-place verbs: the dative codes

the third arguments of verbs of "giving" and "communication" (do, dico),

while the ablative codes the third arguments of verbs of "removing, supply­

ing and depriving" (separo, dono, fraudo). Certain divisions within this

classification scheme are unsupportable. The suggested typology ignores

the various thematic roles coded by "third arguments". Verbs of "remov-