516 ROBERT D. VAN VALIN, JR. & DAVID P. WILKINS
assigned the cognition/propositional attitude relation on the IRH. Thus,
each of the possible interpretations of remember, which is a function of the
semantic content of the y variable in the decomposed semantic structure,
corresponds to a distinct interclausal semantic relation on the IRH. It is
important to note that these same interpretations are found in sentences
with simple NP objects, and this shows that the complement forms are not
the source of the various interpretations; rather, they are a function of
them.
The final stage in the process of relating lexical representations to syn
tactic form is mapping the semantic relations of the IRH into syntactic
clause linkage cagetories and determining the morphosyntactic realization
of these categories in the language in question. The mapping between
semantic and syntactic clause linkage relations is not random, and there are
strong universal constraints on it (cf. FW, pp. 270-2). For example, the
psych-action relation involves a mental disposition to act on the part of the
actor of the complement-taking predicate, and the act is coded in the com
plement. It follows that the actor of the psych-action predicate must also be
an argument of the complement predicate, normally (but not necessarily)
the actor. This means that the two predicates in the linkage must share a
core argument, and this is diagnostic of a core non-subordinate linkage
type. (Cf. "Synopsis", this volume) Thus, a want component in the seman
tic decomposition of the predicate corresponds to a psych-action
interclausal relation and leads to the selection of the appropriate language-
specific complement type which is compatible with the original semantic
interpretation of the predicate. Furthermore, it has been hypothesized
(FW, p. 311) that causative, perception and speech act verbs are the only
types of complement-taking predicates that can contribute two arguments
to a core juncture, and if this is true, then it follows that psych-action verbs
can contribute only one. From this it follows that the nexus type must be
cosubordinate, rather than coordinate. Thus we arrive at the claim that uni
versally the unmarked syntactic linkage type associated with the psych-
action semantic relation is core cosubordination. In English, the normal
realization of this juncture-nexus type is the same-subject infinitive con
struction (see FW, section 6.5), and this is in fact the form that the com
plement of rememberint takes.^2 The instantiation of core cosubordination by
the same-subject infinitive construction is a general morphosyntactic princi
ple of English grammar and need not be stated specifically for remember or
any other psych-action verb. Similar considerations lead to the assignment