Most semantic models are
realist in that they presuppose a more or less unproblematically given
relationship between linguistic items and real-world objects. There is talk of
“the meaning we find in the world about us, both natural and conventional”,
and that the success of communication between two agents A and B depends on
there being “a real situation s such that (i) both A and B know the facts of
s...” (Barwise 1988: 69). But, to repeat an epistemological truism, “only in
the light of interpretation does something become a fact, and only within
processes of interpretation is an observation expressible” (Gadamer, 1989:
30). Scientific inquiry, of course, can only proceed if the existence and
accessibility of the world and its ‘facts’ is pretended to be a non-issue,
and such a view is certainly sound as a working hypothesis if the research
focusses on something unrelated. Semantics, however, is crucially interdependent
with epistemology, because it is often hard to distinguish between meta and
object language: since language is the means of expression for any statement
about the structure of the world and since semantics aims to identify the
structure of meaning, it often seems as if semantics and epistemology are
conflated in one discipline.
This tendency is at least
problematic and most likely fallacious: philosophers have always expressed a
need to explain the existence of the physical world and our mind’s access to
it; cognitive science has shown that it is far from straightforward to explain
how the mind organizes sensory input into ‘objects’. A theory that
presupposes that somebody “knows the facts (of a situation)” is thus built
on sand.
This is not to say, however,
that situation theory is worthless for the semantic project. To the contrary, I
believe that all its import is maintained if it is reformulated in more strictly
linguistic terms by replacing e.g. “situation in the actual world” with
“representation of a (real-world or fictional) situation in discourse”. I
thus reinterpret Barwise and Perry’s system in terms of a discourse
representation theory as initiated by the file change semantics of Kamp (1981)
and Heim (1981).
The denotation of linguistic
items is determined solely within the domain of discourse D, which contains
discourse referents, i.e. those (abstract) entities or sets of entities that are
relevant for the particular utterances at hand. Discourse referents may enter D
via linguistic means (e.g. indefinite descriptions, see Heim 1981), or via
extralinguistic means (e.g. sensory perception). Once in a proper subset of D
that is shared by discourse participants, a discourse referent is communicable,
regardless of its origin. Discourse participants (speaker & listener) may
have different knowledge bases and they may share different subdomains of
discourse with different partners. In other words, they find themselves in
different situations. I thus regard situations as equivalent to proper subsets
of D.
© Philipp Strazny 1998