Advances in Role and Reference Grammar

PREDICTING SYNTAX FROM SEMANTICS^505

properly syntactic aspects of such phenomena as demotion, promotion,
ergativity, voice and surface syntactic relations — all of which it reduces, as
it reduces meaning, to content" (1982:698).
One of the central aims of this paper is to demonstrate the need for,
and to foreshadow the development of, a metalanguage for the decomposi­
tion of predicates which falls between the minimalism of Dowty and the
richness of Wierzbicka. An example of such an intermediate approach is
the method of semantic description in Dixon (1971) which is derived from
the study of the Dyirbal "mother-in-law language". The mother-in-law lan­
guage uses an extremely reduced vocabulary but the same morphosyntax as
the everyday language to convey all of the propositions that can be expres­
sed in everyday Dyirbal itself. Dixon found that non-nuclear predicates
(i.e. more complex predicates) of the same semantic class could all be cap­
tured by a single nuclear predicate in the mother-in-law language, and mod­
ifiers and adjuncts would be used to make more explicit what particular
sense is intended. Dixon maintains that his analysis supports a combination
of a componential and a definitional, natural language approach to seman­
tic description, and it is this type of combined approach that we will
develop. He has not, however, done any subsequent work to develop this
insight.

1.3 Theories of complement selection

Theories of complement choice may be divided into two groups: (1) those

that treat it as a purely syntactic phenomenon, and (2) those that attempt to

relate the choice of a complement to the semantics of the complement-tak­

ing predicate, the semantics of the complement or its complementizer, or

both. The first approach has its roots in the Aspects model, in which the

type of complement(s) a verb takes is listed in its lexical entry as part of its

subcategorization frame, and is further developed in works such as Rosen-

baum (1967). This was the treatment of complement choice in GB until the

appearance of Chomsky (1986) (see above), and it is the basis for the

analyses in LFG and GPSG. In LFG there is a special complement gram­

matical relation, (X)COMP, and it is specified in the subcategorization

frame (which in LFG is stated in terms of grammatical functions instead of

constituents) of the lexical entry of the complement-taking predicate.

GPSG specifies the complements of a predicate in the ID rule introducing

the predicate. In all of these theories, the complements are simply listed,