54 Matthew P. Anstey
Secondly, the FG 4 account is not easily verifiable, as the UR could be
interpreted as containing more referential information than is encoded in
the language. As Harder writes: “If we view ... referents in pragmatic
terms, however, they are merely aspects of the world of discourse after the
utterance, not semantic units which are introduced as entities referred to”
(1992: 320; Harder’s emphases). Not that FG is a stranger to such conun-
drums: witness the semantic neutralization across a single case-marked
noun in Latin or the enormous multifunctionality of genitive constructions
that in many languages reduces to a single morphosyntactic form. Such
phenomena are unfortunately often presented as dichotomous ‘division of
labour’ problems between grammar and pragmatics: is reference better ex-
plained pragmatically or syntactically? But surely the job of reference is
itself shared between syntax and pragmatics. The grammar allows particu-
lar pro-forms to refer to certain types of entities (which is why *Is such a
threat or a promise? is ungrammatical) and the Addressee’s interpretative
strategies choose the appropriate entity from the referential “inventory”
that the pro-form introduces. Lyons captures this balance well when he
writes (1977: 197) that “[t]he linguist can contribute to the study of refer-
ence by describing the grammatical structures and processes which
particular language-systems provide for referring to individuals and groups
of individuals. It does not follow, however, that the linguist must be con-
cerned with the actual reference of expressions in his analysis of the
grammatical structure of system-sentences”.
Thirdly, in defence of FG 4 , we must remember that models such as this
one are not language-specific. Rather, they are designed to represent the
superset of all possible linguistic phenomena. Thus crucial evidence
against this part of FG 4 would be to show that no language has a particular
lexeme used for referring to previous discourse acts. But for a particular
language, we simply need to determine how the language grammaticalizes
referential subtleties. The issue here is to distinguish falsifying FG 4 in gen-
eral from demonstrating that an aspect of FG 4 is irrelevant to a particular
language.
FG 4 is an ambitious formalization of the URs of linguistic expressions.
This formalization in FG 4 has been aligned with the functional hierarchy of
influence, allowing descriptions of linguistic data that represent more
clearly than in FG1-3 the pragmatic, semantic and syntactic factors that in-
teract in language use. Further research is necessary to determine if such
superior descriptions lead to superior explanations.