DISCOURSE REFEREFTS 
Lauri Karttunen 
Universl=y of Texas at Austin 
Department of Linguistics 
Austin, Texas 78712 
Consider an interoretive device that in some manner keeps track 
of individuals that have been mentioned iu a discourse and what has 
been said about them. One feature any such device must have is to he 
able to recognize wPen a novel individual aopears in so~e sentence. 
For examole, in Drocessin~ sentence (1), it N~st recognize that the 
NP a car refers to some yet unmentioned object, which in the following 
sentence may be referred to a~ain by any of tke alternative ways 
in (2). 
(a) It is black. 
(1) I have a car. (2) 
(b) .The car is black. 
However, the same noun phrase does not refer to any car in (3)o 
Consequently, none of the alternatives in (4) is acceptable as a 
continuation of (3). 
(a) *It is black. 
(3) I don't have a car. (I~) u 
(b) *The car is black. 
The above examples show that sometimes the occurrence of an 
indefinite NP in the discourse establishes a 'discourse referent' 
- somethin~ that may'be referred back to by a pronoun or a definite 
noun phrase - and sometimes it does not. There are of course many 
other factors involved besldes negation. For example, (5) is ambiguous 
between two interpretations and only in one sense would it be accept- 
able to continue the discourse with (6). 
(5) Mary wants to marry a Swede. (6) H_ee is tall. 
Several linguists (e.g. LeRcy Baker, Janet Dean) have recently 
studied these constraints on coreferentlallty. They have generally 
investigated the problem in the framework of Noam Chomsky's Aspects 
of the Theory of S.?ntax. It has been assumed that noun phrases are 
associated with 're~rential indices' and marked with respect to the 
feature \[~sDeciflc\],~hat is relevant especially in the case of (5). 
In this paper I will argue that Chomsky's indices and the notion 
of soecificity are not adequate for han~llng the problem of discourse 
referents and that we must adopt the more abstract analysis of noun 
phrases first put forth by Emmon Bach and James D. NcCawley. That is, 
noun phrases must be considered as having an underlying variable which 
is bound by a quantifier similar to those in ordinary ~redlcate 
calculus. In this framework, it is a relatively sinple matter to 
describe the circumstances under which an indefinite NF establishes 
a discourse referent. - This happens in case the sentence bound by 
the ~uantlfler is asserted or presupposed to be true by the speaker. 
~owever, there are some special cases that have to be recognized. 
These involve at least the following: (1) counterfactual conditional, 
(ii) modals, and (iii) quantifler-like adverbs, such as always, 
often, etc. 
