506 ROBERT D. VAN VALIN, JR. & DAVID P. WILKINS
and no attempt is made to predict them from any other factors, e.g. the
semantics of the complement-taking predicate. The non-explanatory
character of this approach was discussed above.
Semantically-oriented approaches to complement selection fall into
two groups. The first takes a semantic characterization of complements into
account but handles the actual selection of complements by particular verbs
in the same way as purely syntactic approaches. Grimshaw(1979), following
Bresnan (1972), divides a subset of English complement types into proposi-
tional, interrogative, and exclamatory, and proposes to add a semantic sub-
categorization frame to the lexical entry of each complement-taking predi
cate to supplement its syntactic subcategorization frame. The second group
tries to relate the selection and form of complements to the semantics of the
matrix predicate, the semantics of the complements themselves, and to the
semantic relation holding between the predicate and its complement.
An early argument for this second approach within the generativist tra
dition can be found in Kiparsky & Kiparsky (1971). They examined factive
and non-factive complement-taking predicates in English in order to show
"that the choice of complement type is in a large measure predictable from
a number of basic semantic factors" (Kiparsky & Kiparky 1971: 345). They
conclude their work as follows:
Syntactic-semantic interrelationships of this kind form the basis of a system
of deep structures and rules which account for the complement system of
English, and other languages as well. The importance of a system success
fully worked out along the general lines suggested above would lie in its
ability to account not only for the syntactic structure of sentential com
plementation, but also for its semantic structure, and for the relationship
between the two. Our analysis of presupposition in the complement system
contributes a substantial instance of the relations between syntax and
semantics, and enables us to correct an error which has been made in most
past work on transformational syntax. The error is that different types of
complements (that-clauses, gerunds, infinitives) have all been assumed to
have the same deep structure, and hence to be semantically equivalent.
We have seen that there is good reason to posit a number of different base
structures, each mapped by transformations into a syntactic paradigm of
semantically equivalent surface structures. ... This approach to a theory of
complementation is not only more adequate from a semantic point of view.
Its purely syntactic advantages are equally significant. It eliminates the need
for marking each verb for compatibility with each surface complement type,
that is, for treating complementation as basically irregular and unpredicta
ble. We account for the selection of complement types quite naturally by our
proposal that there are several meaningful base structures, whose choice is in