Determiners, Entities, and Contexts 
Deborah A. Dahl 
Paoli Research Center 
SDC/A Burroughs Company 
P.O. Box 517 
Paoli, PA 19301 
) 
1. Introduction 
I am concerned with the relationship between the forms of linguistic expressions, noun 
phrases in particular, and the discourse entities to which they refer, s That is, when does a noun 
phrase introduce a new referent into the discourse? My concern in particular is to specify the role 
that the discourse context plays in answering this question. A simple first approach to the rela- 
tionship between noun phrases and discourse entities might suggest that definite noun phrases 
refer to entities which are assumed to be mutually known to the speaker and hearer, and 
indefinite noun phrases refer to entities which are not mutually known, and thus, that discourse 
context plays no role at all. This discussion will point out problems with this approach for both 
definite and indefinite noun phrases. I will describe examples where definite noun phrases are 
used to introduce new referents, and, conversely, where indefnite noun phrases do not introduce 
new referents. In the first case, the local focus structure provides a guide to recognising that a 
new entity is involved, and in the second case, the recognition that no new entity is introduced is 
based on the given/new status of propositions in the discourse. 
I will begin by describing certain definite descriptions that introduce new entities. I will then 
describe some examples where indefinite descriptions do not introduce new entities. In each case, I 
dill discuss some related processing issues. 
I will restrict the current discussion to deal with cases where the mutual knowledge is based 
on the discourse context, rather than on knowledge that the speaker and hearer bring to an 
interaction. In the canes of indefinltes, I will also restrict my discussion to sententlal contexts 
where an indefinite could introduce a new entity; in other words, to specific contexts, as dis- 
tinguished from non-specific contexts as discussed in ~rince1981\]. 
2. Implicit Associates 
The case of definite noun phrases that are intended to introduce new discourse entities has 
been relatively well-researched, in particular by~awkins1978, Hawkins1984\] Hawkins points out 
that entities that have a slot/frame relationship with previously introduced entities often have a 
definite determiner. For example, in 
(1) There were loud noises com;,g from a starting sir compressor. The drive shaft was sheared. 
it is possible to refer to the drive shaft with a definite noun phrase because of its relationship with 
the previously mentioned starting slr compressor, even though the drive shaft has not been men- 
tioned. This same relationship is described by ~rince1981\] as in/errable, and is also discussed in 
~Ieim1982\]. Because we understand the drive shaft mentioned in {1)to be not just any drive 
shaft but the drive shaft that is part of the air compressor mentioned in the previous sentence, a 
full understanding of this noun phrase must capture this relationship. The new noun phrase is 
imp\]ielt\[¥ associated with the local/oeus as described in ~)ahI1986\], and\[Sidner1979\]. 
s The research described in this paper was supported in part by DARPA under contract N000014-85-C-0012, admin- 
istered by the Office of Naval Research, and by • post-doct~rsd fellowship in Copitive Science from the Sloan Founds- 
tion. I have received helpful comments on this paper from John Dowdinc, Lynette Hirechman, Marci• Linebarger, Marth• 
Palmer, Rebecc• SchiS'man, and Bonnie Webber. 
October 1, 1986 154 
In (2d), for example, the referent for the paper seems to be the paper associated with the new 
package even though there is a previously mentioned entity which matches the noun phrase; that 
is, the paper in (2b). 
(2) a. A package arrived yesterday. 
b. The wrapping paper was beautiful. 
c. ~Tnlle I was admiring it, another package arrived. 
d. I removed the paper. 
After the focus change to mlr pae\]:age, the associates of the new package seem to he pre- 
ferred as referents over previously mentioned items, even if the old items had been in focus at one 
time. This is consistent with Sidner's algorithm. ~ 
3. Spec|F;c Attr;butlves 
The second main point to be dealt with in this paper is that of indefinite noun phrases in specific 
contexts, which nevertheless fall to introduce new discourse entities. Most of those who have dis- 
cussed indefinites seem to have assumed that an indefinite reference in a specific context invari- 
ably introduces a new discourse entity. This includes the discussions in ~-\[.Clark1977\] and 
~-\[e\[m1982\]. However, there is a class of indef;nites, which I have called speclJ~c a~trlbu~iees 
~Dah11984~, which I claim do not have this function. 
Consider the example, 
(3) a. Dr. Smith told me that exercise helps. 
h. Since I heard it from a doctor, I'm inclined to believe it. 
An entity, Dr. Smith, is introduced in (3a), and an indefinite noun phrase, a doctor, is used in 
(3b). It is clear that this noun phrase is not intended to introduce a second doctor into the discus- 
s;on. This is an example of a spec~j~c attrlbut~ee. I use the term speeij~c in the sense that a 
specific reference means that the speaker has a particular individual in mind when s/he uses the 
indefinite description. It is clear in (3), for example, that the speaker did not hear that exercise 
helps from some unspecified doctor, but from Dr. Smith. 
The term a~trlbu~iee, as used by ~Donnellan1971\], can also be applied to these indefinites, 
although it was originally suggested only for definites, because the specific identity of Dr. Smith is 
not relevant to the predication, only Dr. Smith's attribute of being a doctor. (See~Dahl1984 .\] for 
detailed arguments about the applicabll\[ty of this term.) 
There are tw~ important issues that must be dealt with in a treatment of specific attribu- 
tives. First, there is the issue of recognizing that the noun phrase in fact is not being used to 
introduce a new entity. Second, it is necessary to recognize the speaker's purpose in using an 
indefinite noun phrase, when a definite noun phrase would have been possible. Both of these issues 
have implications for language generation as well as understanding. For example, in the first case 
a \]~nguage generator w~ have to decide when it is possible to use a specific attributive, and in 
the second case, it wall have to decide whether a specific attributive would be useful in accom- 
plls}~ug its communicative goals. 
I have previously suggested ~Dsb11984~hat a specific attributive can be recognized by its 
occurrence in a proposition that is glwen as in (3), is related to a given proposition by simple 
entailment as in (4), or is related to a given proposition by a plausible inference, as in (5). 
(4) Mary and Bill both volunteered to walk the dog. Since at least one pcrsoft is wi\]l~g to walk 
the dog, we don't have a problem. 
SA discussion by \[Heim19821 suggests that introduction of a new entity with a definite noun phrase is a violation of 
a felicity condition, and is therefore to be handled by a repair or accomodation mechanism. Since accommodation mechan- 
isms are typically triggered by the failure of normal processing, Heim's approach suggests that a failure of normal procees- 
ing would have to occur before • system could recopise that • new referent was being introduced. If normal proee..i.g 
means searching through the discourse context for a referent matching the new description, then the example in (2) pro- 
vides evidence against this position, since the correct processing cannot have been invoked by the failure to find a match- 
in• referent in the previous discourse. 
October 1, 1986 155 
(5) A: I'm afraid I miscalculated Jones's insulin dosage. 
B: What happened? 
A: IIe died. 
B: So, a patient has finally died due to your carelessness. (inference 'Jones is a patient') 
Thus, in order to determine when an indefinite introduces a new entity, it is necessary to 
know whether the proposition in which it occurs is f~ven or nero. For this, we need a represents- 
tion of the events and situations described in the discourse, which can then be examined in order 
to determine when a proposition is given or new. Such a representation, of course, will be needed 
in any case for pronouns or full noun phrases that refer to events and situations. For example, in 
the PUNDIT text processing system, (described in ~a~ner1986\] ), a representation is built for 
each event or situation mentioned. A'noun phrase like the failure in (6) or it in (7) can then he 
recognized as a reference to something previously mentioned. 
(6) The starting air compressor fal\]ed when the oll pressure dropped below 60 psig. The failure 
occurred during the engine start. 
(7) The starting air compressor failed when the oil pressure dropped below 60 psig. I~ occurred 
during the engine start. 
The difference in processing between (6) and (7) on the one hand and specific attributives on the 
other is that for the specific attributives we are saying that something analogous to reference reso- 
lution should be performed on clauses, as we\]\] as on noun phrases. That is, we want to ask 
whether th;~ event has been mentioned before, or can be inferred from something that has been 
mentioned. If so, we can match corresponding participants so that it is possible to recognize that 
no new entity is being introduced, s 
The second issue raised by specific attributives is the speaker's purp()se in selecting an 
indefinite when a definite would have been possible. This seems to he related to the use of 
indefin|tes in genera\] to serve to deemphaslze the particular individual referred to whUe emphasiz- 
ing its genera\] class. In (3), for example, it is not the fact that ~}~im doctor told me that exercise 
would help that is relevant, but rather that the person has the property of being a doctor. Notice 
the contrast between (3) and (8). 
(8) a. Dr. Smith told me that exercise helps. 
h. Since I did hear it from the doctor I'm inclined to believe it. 
(8) suggests that there is something special about Dr. Smith in particular that makes this advice 
reliable, while (3) does not. 
To sum up, I have discussed two categories of noun phrases which demonstrate the effects of 
discourse context on determining whether a new entity is introduced. Implicit associate definites 
introduce new entities which are related to the local focus. Specific attributives refer to prevlous\]y 
introduced entities in given propositions. Minimally, specific attributes have to be recognized, in 
order to prevent the creation of an extra discourse entity, and this requires a representation of 
given propositions. In addition, a complete understanding of specific attributes requires a recogni- 
tion of the speaker's reason for choosing an indefinite when a definite would have been possible. 
References 
~ah\]1984~.'~ 
Deborah A. Dahl, The Rise of Shared Knowledge. Penn Review o/Lingulatlcs 8, 1984, 
pp. 1-14. 
8This raises the issue of what discourse Zosds would be served by repeating something that is already liven. There 
s~e probably a number of' reuons to do this. Investiz~ing them would be u interesting topic for future reselrch. 
October 1, 1986 156 
~Dah11984 k~ 
Deborah A. Dahl, Recognizing Specific Attributives, presented ath the 59th Annual 
Meeting of the Linguistic Society of America, Baltimore, 1984. 
~1988\] 
Deborah A. Dahl, Focusing and Reference Resolution in PUNDIT, Presented at AAAI, 
Philadelphia, PA, 1986. 
~onneUan1971\] 
Keith Donnel\]an, Reference and Definite Descriptions. In Semantlem, D~D. Steinberg 
and L.A. Jakobovits (ed.), Cambridge University Press, Cambridge, 1971. 
~-I.Clark1977\] 
Herbert H.Clark and Eve H.Clark, . In P, ychology ar, I Language, Harcourt, Brace, 
Jovanovich, New York, 1977. 
\[Hawklns1978\] 
John A. Hawkins, Definiteness and IndeJ/n~feneu. Humanities Press, Atlantic High- 
lands, New Jersey, 1978. 
~-Iawkins1984\] 
John A. Hawkins, A Note on Referent Identifiability and Co-Presence. Journn/ o/ 
Pragmat~ea, 1984. 
\[Heim1982\] 
Irene R. Helm, The ..ecmant~es o~ Dej~nlte and lnde.~n~te Noun Phrases. Unpublished 
Ph.D. dissertation, University of Massachusetts, Amherst, \]V~, September 1982. 
\[P*Imer1988\] 
\]V~artha S. Palmer, Deborah ~ Dahl, Rebecca J. SchiEman, Lynette Hirscbman, 1VIarcia 
Linebarger, and John Dowding, Recovering Implicit Information, to be presented at the 
24th Annual Meeting of the Association for Computational Linguistics, Columbia 
University, New York, August 1986. 
~rince1981\] 
Ellen F. Prince, Toward a Taxonomy of Given-New Information. In Radical Pragmat- 
ies, Peter Cole (ed.), Academic Press, New York, 1981. 
\[SidnerI979\] 
Candace Lee Sidner, Towards a Computational Theory of Definite Anaphora 
Comprehension in English Discourse, MIT-AI TR-537, Cambridge, I~L~, 1979. 
October 1, 1986 157 
