514 ROBERT D. VAN VALIN, JR. & DAVID P. WILKINS
make access to the same SRRs with respect to the something.be.in.mind
component, but the actual selection of subclasses is dependent on semantic
compatibility with further elements in the decompositions of the relevant
predicates.
In the SRR representation given above one can see at a glance what
the shared features of the set are and hence what the obvious generalization
for that variable is by comparing the specific decompositions. Further SRRs
would show that know is the only primitive predicate among the set of
decompositions above and that intend is based on the primitive predicate
want, the decomposition of believe would involve the primitive predicate
think (roughly, "x think y be true & χ not know y be true"), and the seman
tic representation of perceive requires know (roughly, "know through the
senses"). As will be seen below, these semantic associations set the founda
tions for the semantic aspects of complements association. Very roughly, if
knowledge is signalled in or consistent with the semantics of the comple
ment taking predicate, a that-complement may be selected; if intentions or
desires are encoded in or consistent with the predicate, a to-complement
may be provided, and so on.
The central fact about remember for this discussion is that all four of
the subtypes of cognitive entities mentioned in the SRR can fulfill the role
of something.be.in.mind.from.before; that is, the entity filling the y vari
able can be an intention, some piece of knowledge, a belief, or a percep
tion. For ease of reference only, these interpretations of remember will be
labelled "remember'int", "rrknw^\ " rememberblf ,and "remember ",
respectively.
We now turn to the question of how to relate these semantic represen
tations to the syntactic complement types that are possible with remember.
The key to this is the RRG Interclausal Relations Hierarchy [IRH], the
latest version of which is presented in Figure 1. In the IRH, the semantic
relations holding between the units in a complex sentence are correlated
with the syntactic type of the linkage. The semantic relations do not as yet
have full explications of their content, and one of the tasks for future
research is the development such explications, using the same semantic
metalanguage employed for the decomposition of verbs. Nevertheless,
rough approximations of their content will be adequate for the purposes of
this discussion. The semantic relations relevant to the analysis of remember
are psych-action, direct perception, and cognition/propositional attitude.
"Psych-action" may provisionally be characterized as signalling some kind