FG from its inception 53to make one academic discipline have evaluative authority over another.
Even when psychology does bolster our case, we should beware of saying
that our model has ‘psychological adequacy’. Perhaps we could say instead
that the symbolic meaning we attribute to our model coheres well with
psycholinguistic evidence.
The last issue that needs to concern us here is PR7, that is, the problem
of formalization. We are now in a position to evaluate the core of FG 4 : its
formal structure. How does it line up with linguistic data? Irrespective of
the hypothetical pathway taken by the Speaker in deriving URs, the nota-
tion can only be justified if it provides superior linguistic explanations.
We should observe from the outset that the greater part of Hengeveld’s
article is a justification of an extension of the formal notation to capture
many subtle referential problems that “would have been difficult to handle
in earlier versions of FG” (this volume: 17). Hengeveld is therefore moti-
vated by problems of actual language data (hedged performatives, identity
statements, metaconditionals, and direct speech). Consequently, a crucial
question regarding FG 4 is, in my opinion, whether the capacity of the
model to describe referentiality exceeds the capacity of grammar. Rather
than make a definitive judgement on this question, I wish to raise three ba-
sic issues that need to be considered.
Firstly, Hengeveld clearly sides with the view that the URs are bearers
of meaning: “the speaker refers to entities”, “the speaker ... ascribes prop-
erties to entities” (this volume: 6), and since most of his problems are refer-
ential he has invested the model with referential precision. Whatever a pro-
form refers to, Hengeveld needs a variable for it. So let us take Hengeeld’s
example (9) again, now numbered (27), and pretend B replied differently:
(27) a. If you behave well, I’ll let you read my poems.
b. Is that Keats or Donne?
In this case the demonstrative that would refer to ‘my poems’. This in-
dicates that the expression rules must select the same pro-form, that, to
represent both types of anaphora. Thus FG 4 , based on URs as meaning-
bearers, models a process of referential neutralization whereby a single
pro-form does many referential duties. An alternative model, based on a
pragmatic account of reference, would have the Speaker select that to trig-
ger an interpretative strategy whereby the Addressee determines the most
appropriate antecedent (as suggested by Harder 1996a: 237–243). The
point is clear: the solution to the problem of URs (PR2) affects how one
chooses between these alternatives.