<< >> TOC
The insistence on using discourse representations to
identify meanings without any recourse to the ‘real world’ is not just an
attempt to dodge epistemological problems, but has some – I believe –
welcome consequences for the evaluation of proper names. One question that may
immediately come to mind is how typical instances of de re vs. de dicto readings
may be explained, if the real world is not accessible to yield a de re referent?
Consider the following example (from Lerner/Zimmermann 1991, henceforth L/Z):
(b) Werner is identical to Erwin.
(c) Monika does not believe that Erwin broke his foot.
In the de dicto reading, the interpretation of Werner and
Erwin are taken to be interpreted within the scope of the propositional attitude
verb (PAV) believe. The denotation of both Werner and Erwin may thus be
determined within a belief-context. If (1b) is not part of the same
belief-context, (1a) and (1c) are not contradictory.
The de re
reading in a realist model interprets both occurrences of proper names against
‘reality’ as backdrop. Since (1b) is ‘part of reality’, (1a) and (1c)
have to be regarded as contradictory. L/Z propose to incorporate an access
relation into the definition of de re interpretations of propositional attitude
contexts. This access relation is supposed to hold between the subject of the
propositional attitude verb and an object within the scope of the PAV. Thus,
(the actual) Monika has access to (the actual) Werner via two names, and – if
independent from each other - these access relations allow independent beliefs
that may differ without causing a contradiction.
In essence, I
agree with L/Zs position. However, the whole problem complex only arises in a
realist model. I believe that a non-realist interpretation of discourse
representations that takes the inescapability of discourse into account yields a
solution that is even more straightforward.
In a strictly
text centered analysis, neither an ‘actual Monika’ nor an ‘actual
Werner/Erwin’ are relevant. Both Monika and Werner/Erwin are discourse
referents mentioned in the exchange between the authors L/Z and the readers
(us). Thus, in the notation used by Heim (1981), the relevant discourse
representation shared by L/Z and the reader is the following:
There are thus
two discourse referents x and y; x has one name, y has two. Inhowfar the
two discourse referents have a relation to actual people in the real world is
completely irrelevant. Sentence (1b) is a communication between L/Z and the
reader and is thus reflected in the fact that Werner and Erwin are conflated in
one discourse referent. Sentences (1a) and (1c) add to the above picture in that
the discourse referent x, i.e. Monika is supplied with her own subdomain of D,
as shown below:
The readers are told that the discourse
referent Monika has beliefs, i.e. it must have access to a discourse
domain,which is – in this case – a subdomain of D (the domain shared by L/Z
and the reader), since D contains all information about this domain. This
subdomain of D contains two further discourse referents, one of them Werner, the
L/Z formulate a pragmatic constraint on name usage in
communication between cooperative speakers:
(4) Don’t name the same thing in different ways within the same context!
This constraint is at work in the communicative
framework between L/Z and the reader: even though L/Z never make it explicit,
they obviously assume that the discourse referents in DMonika are identical in
reference to the discourse referent y in DL/Z. However, this may be so, or it
may not. The important point is that the linguistically relevant items are
discourse referents (as opposed to ‘people’), and in the case at hand, there
are four of them now: X/Monika, Y/Werner/Erwin, U/Werner, and V/Erwin.
(1a) and (1b) occur within the communicative framework DL/Z. The items Werner
and Erwin occur within the scope of Monika believe, i.e. they are element of
DMonika, and since DMonika ' DL/Z, they are accessible for interpetive purposes.
An occurrence of e.g. Werner thus allows to be interpreted within DMonika or
within DL/Z. The first yields the de dicto interpretation, the second one the
so-called de re interpretation, which simply means that discourse participants
sharing DL/Z follow constraint (4) and conflate e.g. Werner (' DL/Z) with Werner
dicourse representation theory allows to capture the difference between de dicto
and de re interpretation without recourse to either realist assumptions (in the
case of de re) nor to substitutions of definite descriptions for names (the
Russellian treatment of de dicto cases). The difference interpretation is merely
on of difference in context, i.e. the (sub)domains of discourse used for
interpretation. Such subdomains can be freely defined as being associated with a
particular discourse referent (as in 3), or the main discourse domain can be
subdivided itself. This would be the case when fictional discourse referents are
used, e.g. Miss Marple:
(5) Miss Marple outsmarts the villain.
In (5), the subdomain D: Miss Marple stories is supplied
pragmatically, provided the listener knows that Miss Marple is a fictional
character or can infer this fact from the use of the villain. Of course, the
subdomain can also be introduced overtly, e.g. with introducing phrases such as
In Agatha Christie’s novels....
representation theory, typical realist issues – e.g. whether or not a name
refers to a fictional character or whether or not the referent of a name is
“properly” identifiable for a particular discourse participant – become
unproblematic and straightforward to represent. With Heim, I assume that
“logical forms are not assigned truth conditions, only [discourse
representations] are” (1981: 9a). Under this assumption, realist implications
become important in the evaluation of discourse representations, but are
completely removed from the strictly linguistic level. This is, of course, a
welcome result, since language – as a computational system – handles e.g.
names of “fictional characters” just as well and in exactly the same manner
as names of “real people”.
With the backing of discourse representation theory, I will
thus ignore the more philosophically relevant issues of reference and
concentrate on the strictly linguistic characteristics of proper names. Three
problem complexes have to be dealt with here:
(A) type assignment, semantic (individual vs. set)
and syntactic (N vs. DP)
(B) semantic function (indefinite vs. definite)
(C) syntactic distribution
These problem complexes are, of course, interdependent:
semantic types are the basis for determining the semantic function; under the
assumption that syntax encodes semantic structure, semantic functionality should
find expression in the syntactic distribution; syntactic distribution is based
on syntactic type, which again should be compatible with semantic type
assignment. In my inquiry, I will start with the issue of semantic type and
roughly follow the circular path described above in order to finally come to
compatible assessments of the semantic and syntactic structure of proper names.
<< >> TOC
© Philipp Strazny 1998