A SYNOPSIS OF ROLE AND REFERENCE GRAMMAR 83therefore the NP in the PCS must be the one which would function as pivot
in a non-WH-question. Thus the linking of the PCS WH-NP and the theme
argument position in the LS is supported by a number of considerations.
This linking is represented in Figure 20.
It must be emphasized that even though these linkings are described in
terms of a sequence of steps, they are not equivalent to the stages in a trans
formational derivation, for two reasons. First, there is only one syntactic
representation, the sentence itself, and one semantic representation, the LS
of the predicate. Consequently, there are no derivationally related syntactic
representations which are mapped into each other. Second, each step
involves interpreting the structure in question, not manipulating or trans
forming it in any way. The argument positions in LS are interpreted to yield
the thematic relations of the arguments, and then these are assigned mac-
rorole status (or not) with respect to the hierarchy in (25). The output of
step 2 in (63) is not a distinct level of representation but rather the accumu
lation of this semantic information about the arguments. Maria in (64b) is
the first argument of do' in the LS for present, and the result of step 2 is that
this characterization is enriched to include the information that this argu
ment is an effector and the actor. Thus (65) is equivalent to an information-
ally enriched version of (64b), not a distinct level of semantic or syntactic
representation.
The linking algorithms in (63) and (70) are based largely on the
analysis of English, but their basic structure is universal, as represented in
Figure 16. Only through the detailed analysis of specific languages of vary
ing types will a formal statement of the universal aspects of linking emerge
out of the mass of language-particular details, for it is only through such