Predicting Syntactic Structure
from Semantic Representations:
Remember in English and its Equivalents in Mparntwe ArrernteRobert D. Van Valin, Jr. & David P. Wilkins
SUNY Buffalo- Introduction
The aim of this paper is to demonstrate the means by which the syntactic
properties of a predicate can be predicted from its semantics within the
Role and Reference Grammar [RRG] framework. More specifically, our
goal is to show how the types and form of the complements that a predicate
takes can be deduced directly from the semantic representation in its lexical
entry. The result will be an explanatory account in which the lexical entry
for a complement-taking predicate will contain only its semantic represen
tation; the syntactic properties of its complements will be derived from that
representation together with a set of independently-motivated, language-
specific and universal semantic, lexical and morphosyntactic principles. To
this end, we will investigate the complement-taking predicate remember in
two typologically distinct languages, English and Mparntwe Arrernte
(Aranda), an Australian Aboriginal language of Central Australia. We
have chosen remember because of the rich set of complement forms and
associated meanings that it takes in English {John remembered to close the
door, John remembered that he had closed the door, John remembered clos
ing the door); this presents a challenge with respect to its analysis in English,
and it raises interesting comparative questions: how is this semantic domain
carved up in Mparntwe Arrernte? is there a single verb covering all of these
meanings? does the same correlation between form and meaning hold? We
will develop a system of semantic decomposition for predicates, starting