PREDICTING SYNTAX FROM SEMANTICS 507large part predictable from the meaning of each predicate. (Kiparsky &
Kiparsky 1971: 365; emphasis ours)
While this basic orientation has largely been abondoned by those in the
generativist tradition, it has been sustained by functionalists. The RRG
theory of clause linkage falls into this group (see section 2), as do the
approaches presented in Silverstein (1976,1980,1981), and Givón (1980).
Both Silverstein and Givón present hierarchies of cross-clause semantic
relations and argue that the choice of complement is a function of these
semantic relations in terms of the following correlation: the closer (i.e.
higher on the hierarchy) the semantic relation between the two clauses is,
the tighter the syntactic linkage between them will be and the greater the
formal deformation of the dependent clause from the unmarked main
clause form. The RRG theory of clause linkage is in fact based on Silvers-
tein's proposals.
There are four other approaches which attempt to correlate the syntax
and semantics of complementation. Dixon (1984) proposes that it is possi
ble to predict the complement(s) a verb may take based on the following
facts: (1) the kinds of complement clause(s) the language operates with; (2)
the semantic type to which a verb belongs and its further specification
within that type; (3) the complement possibilities for that semantic type in
that language; and (4) relevant facts about the syntactic organization of the
language, i.e. its possible syntactic structures and constraints (1984:585).
His paper deals with only (1) and (3), and, while his basic claim is an impor
tant one, one which is very close to the position advocated in this paper, it
is not embedded in any theory of syntax or semantics and therefore remains
an intriguing but ad hoc set of observations. Ransom (1986) investigates the
interaction of the "modalities" of complement clauses ("information mod
alities", e.g. truth, and "evaluation modalities", e.g. predetermined) with
those of the higher clauses in the determination of complement form, while
Rudanko (1989) presents a classification of English complementation and
control patterns, using a case grammar framework. Both of these analyses
assume the same syntactic structures proposed in Aspects-style transforma
tional grammar; in particular, they both take infinitival complements to be
the direct object of the verb, a position argued against in FVV, section
6.2.1. Ransom's account uses different semantic and pragmatic factors from
those discussed in this paper and therefore is complementary to the analysis
to be presented here. If the type of analysis proposed here is successful, it
would provide an explanation for the classification that Rudanko puts
forth.