Referring as Requesting

(1)

R e f e r r i n g a s R e q u e s t i n g

Philip R. Cohen

Artificial Intelligence Center SRI International

and

C e n t e r for the Study of Language and Information Stanford University

1. I n t r o d u c t i o n 1

Searle [14] has a r g , e d forcefully t h a t referring is a speech act; that people refer, uot just expressions. This paper considers what kind of speech act referring might be. I propose a generalization of Searle's "propositional" act of referring t h a t t r e a t s it as an illocutionary act, a request, and argue t h a t the propositional act of referring is unnecessary.

The essence of the a r g u m e n t is as follows: First, I consider Searle's definition of the propositional act of referring {which I t e r m the PAA, for Propositional Act Account). This definition is found inadequate to deal with various utterances in discourse used for the sole purpose of referring. Although the relevance of such utterances to the propositional act has been defined away by Searle, it is clear t h a t any comprehensive account of referring should t r e a t them. l develop an account of their use in terms of a speaker's

requesting

the act of referent identification, which is to be understood in a perceptual sense. This

itlocutionary act analysis

(IAA) is shown to satisfy Searle's conditions for referring yet captures utterances t h a t the PAA cannot. The con- verse positif)n is then examined: Can the IAA capture the same uses of referring expressions as the PAA? If one extends the perceptually- based notion of referent identification to include Searle's concept of identification, then by associating a complex propositional a t t i t u d e to one use of the definite determiner, a request can be derived. The ]AA thus handles the referring use of definite noun phrases with independently motivated rules. Referring becomes a kind of requesting. llence, the propositional act of referring is unnecessary.

2. R e f e r r i n g a s a p r o p o s i t i o n a l s p e e c h a c t

Revising Austin's [2] I o c u t i o n a r y / i l l o c u t i o n a r y dichotomy, Searle distinguishes between illoeutionary acts (IAs) and propositional acts (PAs) of referring and predicating. Both kinds of acts are performed in making an utterance, b u t propositional acts can only be performed in the course of performing some illocutionary act.

Let us consider Searle's rules for referring, which I t e r m the "propositional act analysis", or PAA. A speaker, S, "successfully a n d non-defectively performs the speech a c t of singular identifying refer- ence" in uttering a referring expression, R, in the presence of hearer, H, in a context, C, if and only if:

1. Normal input and o u t p u t conditions obtain. 2. Tile utterance of R occurs as part of the u t t e r a n c e of

some sentence (or similar stretch of discourse) T. 3. The utterance of T is the (purported) performance

of an illocutionary act.

4. There exists some object X such t h a t either R con- ]The research reported in this paper was supported initially by the Fairchild Cam-

era and Instrument Corp. Its subsequent development has been made possible by a gift from the System Development Foundation. I have benefitted from many discussions with Itector Levesque and Ray Ferra.ult.

rains an identifying description of X c r S is able to supplement R with an identifying description of X. 5. S intends t h a t the utterance of R will pick out or

identify X to If.

6. S intends t h a t the utterance of R will ideatify X to tt by means of If's recognition of S's intention to identify X and he intends this recognition to be achieved by means of It's knowledge of the rules governing R and his awareness of C.

7. The semantical rules governing R are such t h a t it is correctly uttered in T in C if and only if conditi~ms 1-6 obtain." (

l]4],

PP. 94-95.)

Conditions 2 and 3 are justified as follows:

Propositional acts cannot occur alone: t h a t is oue cannot

just

[emphasis in original - - PRC] refer and pred- icate without m a k i n g an assertifm or asking a question or performing some other illocutionary act . . . One only refers as part of the performance of an illocutionary act, and the g r a m m a t i c a l clothing of an iilocutionary act is the complete sentence. An utterance of a referring expression only counts as referring if one says something.

(Ibid,

p. 25.)

The essence of Conditions 4 and 5 is t h a t the speaker needs to u t t e r an "identifying description". For Searle, "i(lentification ~ means "... there should no longer be any doubt what exactly is being talked about".

(lbid,

p. 85.) Furthermore, not only should the description be an identifying one (one t h a t would pick out an object), but the speaker should intend it to do so uniquely (Colidition 5). Moreover, the speaker's intention is supposed to be recognized by the hearer {Condition 6}. This l a t t e r Gricean [7] condition is needed to distin- guish having the hearer pick out an object by referring to it versus, for example, hitting him in the back with it.

3. P r o b l e m s f o r t h e P r o p o ~ i t i o n a l A c t A c -

c o u n t

In a recent experiment [3], it was shown t h a t in giving instructions over a telephone, speakers, but not users of keyboards, often made s e p a r a t e u t t e r a n c e s for reference and for predication. Frequently° thes.e "referential utterances" took the form of existent ial sentences, such as "Now, there's a black O-ring". Occasionally, speakers used question noun phrases "OK, now, the smallest of the red pieces?" The d a t a present two problems for the PAA.

3.1. Referring as a Sentential Phenomenon

(2)

s o m e t h i n g . f the refi, rent as p a r t of the p e r f o r m a n c e of some illocu- t i , n a r y act. llowever, it is ~bvious t h a t speakers c a n refer by issuing isolated noun phra,:e or p r e p o s i t i o n a l phrases. Since s p e a k e r s performed illoculi,mary a c t s iu m a k i n g these u t t e r a n c e s , then, a c c o r d i n g to C o n d i t i . n s 2 a n d 3, tl'ere should be an a c t of predication, e i t h e r in the sentence . r the "similar s t r e t c h of discourse". For e x a m p l e , consider the f . l ! , w i n g dial,,gue fragment:

1. "Now, the small blue c a p we talked a b o u t before?"

'2. "l~'h-h u h",

3. " P u t that over the hole Oil the side of t h a i tube . . . "

The i l l o c u t l o n a r y a r t p e r f o r m e d by u t t e r i n g p h r a s e ( i t is finished a n d responded to in p h r a s e (2} before the i l l o c u t i o n a r y a c t p e r f o r m e d in p h r a s e (3) c o n t a i n i n g the p r e d i c a t i o n " p u t " is p e r f o r m e d . T h e ap- peal to a sentence or s t r e t c h of discourse in which t o find the illocu- t i o u a r y a c t c o n t a i n i n g tile p r o p o s i t i o n a l a c t in (1) is therefore is un- convincing. T h e cause of this i n a d e q u a c y is t h a t , a c c o r d i n g to Searle, to p e r f o r m an i l l . e u t i o n a r y a c t , a n a c t of p r e d i c a t i n g is required, a n d the p r e d i c a t e m u s t be u t t e r e d (Ibid, pp. 1k26.-127~. llence, there is no a p p e a l to c o n t e x t to supply t,bvious predications. Likewise, there is no r o o m for c o n t e x t to supply a n obvi~ms focus of a t t e n t i o n . U n f o r t u - nately, we can easily i m a g i u e cases in which a n object is mutually, but nonlinguistically, b c u s e d upon {e.g., when Ilolmes, h a v i n g c o m e u p o n a b o d y on the g r o u n d , listens for a h e a r t b e a t , a n d says to W a t s o n : " D e a d " ) . In snch a case, we need only l)redicate. T h u s , the require- ment t h a t the a c t of reference b~" j~,intly l o c a t e d with some p r e d i c a t i o n in a sentence or ilh~cutiuuary a c t is t~o restrictive - - the 9oals involved with reference a n d p r e d i c a t i o n c a n be satisfied s e p a r a t e l y a n d c o n t e x - tually. T h e point of this p a p e r is to bring such goals to the fore.

3.2. Referring w i t h o u t a P r o p o s i t i o n a l Act

Tile second pr(llqi'ni is t h a t Inost of tile s e p a r a t e u t t e r a n c e s issued to secure reference were declarative sentences whose logical form was 3 z P(z). For e x a m p l e , " t h e r e is a little yellow piece of r u b b e r " , a n d "it's got a plug in it". However, Searle claims t h a t these u t t e r a n c e s contain no referring act. (lbid, p. 29.) flow then c a n s p e a k e r s use them to refer?

T h e a n s w e r inwdves a n a n a l y s i s of indirect speech acts. A l t h o u g h such declarative u t t e r a n c e s can be issued just to be informative, they are also issued as requests t h a t the h e a r e r identify the referent. : T h e analysis of these u t t e r a n c e s as requests d e p e n d s on o u r positing a n a c t i o n of referent identification.

4 .

I d e n t i f i c a t i o n a s a R e q u e s t e d A c t i o n

In Searle's a c c o u n t , speakers identify referents for hearers. ! re- vise this notion slightly a n d t r e a t identification as an act p e r f o r m e d by the h e a r e r 131 . I use the t e r m "identify" in a very. n a r r o w , t h o u g h i m p o r t a n t a n d basi-, sense - - one t h a t i n t i m a t e l y involves p e r c e p t i o n . Thus, the a n a l y s i s is not intended to be general; it applies only to the case when the referents are p e r c e p t u a l l y accessible to the h e a r e r , a n d when the h e a r e r is intended to use p e r c e p t u a l m e a n s to pick t h e m out. For the time b~'iug. I a m explicitly not c o n c e r n e d with a h e a r e r ' s mentally "identifying" some entity satisfying a description, or discov- ering a coreferring descripticm. T h , p e r c e p t u a l use of "identification" would a p p e a r to be a ~pe('ial case of .%arb"s use of the term, a n d thus

Searle's condition,s sh,.uhl apply to, it.

Referent identifica~:ion in this p e r c e p t u a l sense requires a n agent and a description. T h e essence of the a c t is t h a t the agent pick out the t h i n g o r t h ! n f f s satisfying the description. T h e a g e n t need not '2The classification of these utterances as identification requests was done by two coders who were trained by the author, but who worked independently. The reliability of their ¢odings were high - - over 90 per cent for such existential ~tatements.

be the s p e a k e r of the description, a n d indeed, the description need not be c o m m u n i c a t e d linguistically, o r even c o n l m u n i c a t e d a t all. A crucial c o m p o n e n t of referent identification is the a c t of p e r c e p t u a l l y searching for s o m e t h i n g t h a t satisfies the description. T h e description is d e c o m p o s e d by the a g e n t into a plan of a c t i o n for identifying the referent. T h e intended a n d e x p e c t e d physical, sensory, a n d cognitive a c t i o n s m a y be signalled by tile s p e a k e r ' s choice of predicates. For e x a m p l e , a s p e a k e r u t t e r i n g the p h r a s e " t h e m a g n e t i c s c r e w d r i v e r " , m a y intend for the h e a r e r to place various s c r e w d r i v e r s a g a i n s t some piece of iron to d e t e r m i n e which is m a g n e t i c . Speakers know t h a t h e a r e r s m a p (at least some) predicates, o n t o a c t i o n s t h a t d e t e r m i n e their extensions, a n d thus, using a model of the h e a t e r ' s capabilities a n d the c a u s a l c o n n e c t i o n s a m o n g people, their senses, a n d physical objects, design their expression, D, so t h a t h e a r e r s c a n successfully e x e c u t e those a c t i o n s in the c o n t e x t of t h e overall plan.

Not only does a s p e a k e r plan for a h e a r e r to identify the referent of a description, b u t he often indicates his intention t h a t the h e a r e r do so. A c c o r d i n g to Searle, one possible way to do this is to use a definite d e t e r m i n e r . O f course, not all definite N P ' s are used to refer: for e x a m p l e , in the sentence " t h e last piece is the n o z z l e ' , the referent of the first N P is intended to be identified, w h e r e a s the referent of the second N P is not. T h e a t t r i b u t i v e use of definite noun p h r a s e s [6] is a case in which the s p e a k e r has no intention t h a t the h e a r e r identify a referent. Yet o t h e r n o n a n a p h o r i c u~es of definite noun phrases include labeling an object, c o r r e c t i n g a referential m i s c o m m u n i c a t i o u , h a v i n g the h e a r e r wait while the s p e a k e r identifies the referent, etc. '~

To r e s p o n d a p p r o p r i a t e l y , a h e a r e r decide~ when identi[icatiou is the act he is s u p p o s e d to p e r f o r m on a description, w h a t p a r t this a c t " will play in the s p e a k e r ' s a n d h e a r e r ' s plan~, a n d how a u d when to p e r f o r m the act. If p e r c e p t u a l l y identifying a feb'rent L~ represented as a n a c t i o n in the s p e a k e r ' s plan, hearers can reas,,u ab<,ut it just as a n y o t h e r act. t h e r e b y allowing them to infer the s p e a k e r ' s iat;'ntions behind indirect identification re.quests.

In s n m m a r y , referent identification shall mean l.lie c o n d u c t i n g =,f a p e r c e p t u a l search process for the referent <,f a description. Tile verb

"pick o u t " should be taken as s y n o n y m o u s .

T h e following is a sketchy definition of tlw ceil,rent identification action, in which the description is formed fr~mi " a / l h , • y such t h a t D ( y ) " . 4

3 X [ P E R C E P T U A L L Y -

A C C E S S I B L E ( X . A g t ) & D ( X ) &

I D E N T I F I A D L E ( A g t , D) D

R E S U L T ( A g t ,

I D E N T I F Y - R E F E R E N T { A g t . D). I D E N T I F I E D - R E F E I'HDNT

(Agt. D. X)I

T h e f o r m u l a follows the usual axi,ml;ttizati,m of acti,)n~ in dynamic logic: P D fAct]Q; t h a t is, if P, aft,'r d . i u g Act. Q. ["..ll.,'*ing M o o r e ' s [9] possible worlds s e m a n t i c s for actbJn, tit," nv,,,lal o p e r a u , r R E S U L T is taken to be true of a n a g e n t , an a c t i . n , an,I a formula, iff in the world resulting from the a g e n t ' s perf,~rmfltg t h a t actitm, tl,e f o r m u l a is true. s

T h e a n t e c e d e n t includes three conditions. The first is a -per- 3See also {15, IO] for discu-:~ion of sl)~,akors' goal~ toward~ th,. irC~r i r,.,a'i,,n of

descriptions.

4Thls definition is not particularly ilhlullnating, t, ut it is ng!. ;lily ,:a'.:u,.r thza Searle's. The point of giving it is that if a definition can be given in this form (i.e., as an action characterizable in a dynamic logic), the illocuti~nary analysis applies.

SActually, Moore characterizes R.ESULT as taking an event and a fornmla as

(3)

ceptual accessibility" condition to guarantee t h a t the I D E N T I F Y - R E F E R E N T action is applicable. This should guarantee t h a t , for example, a speaker does not intend someone to pick out the referent of "3 ~, "democracy", or "the first m a n to land on Mars ~. The condition is satisfied in the e x p e r i m e n t a l task since it rapidly becomes m u t u a l knowledge t h a t the task requires communication a b o u t the objects in front of the hearer.

The second condition s t a t e s t h a t X fulfills the description D. Here, 1 am ignoring cases in which the description is not literally true of the intended referent, including metonymy, irony, and the like (but see [12]). Lastly, D should be a description t h a t is identifiable to this particular A g t . It should use descriptors whose extension the agent already knows or can discover via action. I am assuming t h a t we can know that a combination of descriptors is identifiable without having formed a plan for identifying the referent.

To give a name to the s t a t e of knowledge we are in after having identified the referent of D, we will use { I D E N T I F I E D - R E F E R E N T A g t D X ) . T h a t is, A g t has identified the referent of D to be X. Of course, what has been notoriously difficult to specify is just what A g t has to know a b o u t X to say he has identified it as the reforent of D. At a minimum, the notion of identification needs to be made relative to a purpose, which, perhaps, could be derived from the bodily a c t i . n s t h a t someone (in the context} is intended to perform upon X. Clearly, "knowing who the D is" 18, 9], is no s u b s t i t u t e for having identified a referent. After having picked out the referent of a doseription D, we may still not not know who the D is. On the other hand, we may know who or what the description denotes, for example, hy knowing ~ome "standard name" for it, and yet be unable to use t h a t kn.wledge to pick out the object. For example, if we ask "Which is tho Seattle train?" and receive the reply "It's train number 11689", we may ~till not be able to pick out and board the train if its serial number is not plainly in view.

Finally, athough not stated in this definition, the means by which the act is performed is some function m a p p i n g D to some procedure that, when executed by A g t , enables A g t to discover the X t h a t is the referent of D.

4 . 1 . R e q u e s t i n g

Con.~ider what it takes to make a request. Hector Levesque and I [4, 5] argu~, that requests and other illocutionary acts can be defined in terms of interacting plans - - i.e., as beliefs a b o u t the conversants' shared knowledge of the speaker's goals and the causal consequences of achieving those goals. In this formalism, illocutionacy acts are no longer conceptually primitive, but r a t h e r a m o u n t to theorems t h a t can be proven a b o u t a s t a t e of affairs. T h e proof requires an a x i o m a t i z a - tion of agents' beliefs, expectations, goals, plans, actions, and a cur* relation of utterance mood with certain propositional attitudes. The i m p o r t a n t point here is t h a t the definition of a request is not merely stipulated, but is derived from an independently m o t i v a t e d theory of action. Any act t h a t brings a b o u t the right effects in the right way satisfies the request theorem.

Briefly, a request is an action (or collection of actions) t h a t makes it ( 1 ) shared knowledge t h a t the speaker's goal is t h a t the hearer thinks the speaker wants the hearer to a d o p t the goal of doing a p a r t i c u l a r act, thereby m a k i n g it (2) shared knowledge between the speaker and hearer t h a t the speaker wants the hearer to do t h a t act. This inference requires an additional "gating" condition t h a t it be shared knowledge t h a t the speaker is b o t h sincere and can perform the requested a c t (i.e., he knows how, and the preconditions of t h a t act are true).

The processing of an utterance is assumed to begin by applying the propositional a t t i t u d e correlated with its mood to the propositional content associated with its literal interpretation. Thus, correlated with imperatives and interrogatives is the a t t i t u d e above {corre-

spending to goal (I)): 6

( M U T U A L - B E L I E F H e a r e r S p e a k e r ( G O A L S p e a k e r

( B E L H e a r e r ( G O A L S p e a k e r

( G O A L H e a r e r

( D O N E H e a r e r A c t P ) ) ) ) ) } .

( D O N E H e a r e r A c t P ) is true if H e a r e r has done act A c t and has brought a b o u t P . For y e s / n o interrogatives, A c t would be an I N F O R M I F [11]; for imperatives, it would be the act mentioned in the sentence. Declaratives would be correlated with a different propositional attitude. Beginning with the utterance-correlated attitudes, a derivation process t h a t constitutes plan-recogniti.n reasoning determines what the speaker meant [7]. Thus, for example, what the speaker meant could be classified as a request if the derivation included m a k i n g (2) true by m a k i n g the above formula true.

An act may simultaneousJy achieve the goals constituting more than one illocutionary act. This ability underlies the analysis of indirect speech acts. Formalisms have been developed [5, 11] t h a t describe when we can conclude, from a speaker's wanting the hearer to want the precondition of some act to hold, (or wanting the hearer to believe the precondition does hold), t h a t the speaker wants the hearer to adopt the goal of performing the act. The c o n d i t i . n s licensing this inference are t h a t it be mutually known t h a t the act (or its effect) is an expected goal of the speaker, and t h a t it be mutually known that: the hearer can perform the act. is cooperative, and does not want not to do it.

Returning to the troublesome existential sentences, this pattern of reasoning, which I term the "illoculionary act analysis" {[AAI, can be used to derive a request for reh'rent identification. The reasoning is similar to that needed to infer a request to open the door on hearing a speaker, with two arm-loads of groceries, say "the door i~ closed". The general form of this reasoning involves the a.~serti.n o! an action's precondition when the effect of the action is an expected goal of the speaker. In the case a t hand, the speaker's existential assertion causes the hearer to believe the existential precondition of the referent identification act, since speaker and hearer both think they are talking a b o u t objects in front of the h~'arer, and because the description is identifiable. Ilence, the hearer concludes he is intended to pick out its referent. The hearer may go on to infer t h a t he is intended to perform other acts, such as to pick up the object. This inference process also indicates when the indirect request interpretation is not iutended, for example, if it is mutually known t h a t the description is not identifiable, or if it is m u t u a l l y known t h a t the hearer would not want to identify the referent.

I argue t h a t this kind of reasoning underlies the propositional act account. First, I show t h a t Searle's conditions on referring are a special case of the conditions for requesting referent identification. Then, I show t h a t if one e x t e n d s the definition of I D E N T I F Y - R E F E R E N T to cover Searle's more general concept of identificatiou, the IAA is applicable in the s a m e circumstances a~s Searle's analysis. Because the IAA is independently m o t i v a t e d and covers more cases, it should be preferred.

5. A c c o u n t i n g for Searle's C o n d i t i o n s on

R e f e r r i n g

(4)

requested action (i.e., I D E N T I F I E D - R E F E R E N T } by way of c o m m u - n i c a t i n g (in the (;ricean sense) his intent t h a t the h e a r e r p e r f o r m the action, provided t h a t it is s h a r e d knowledge t h a t the h e a r e r c a n d o the a c t i . n . T h e last condition is true if it is s h a r e d knowledge t h a t the the p r e c . n d i l i . n to the a c t i o n hohts, which includes Seacle's existential Condition 4. Scarle's C o n d i t i o n 5 s t a t e s t h a t the s p e a k e r intends to identify Ihe referent to the hearer. This condition is c a p t u r e d in the IAA by its b e e . r u i n g u m t u a l knowledge t h a t the s p e a k e r intends to achieve the effect of the referent identification a c t , I D E N T I F I E D - IH'H"I';RENT. Finally, Searle's G r i c e a n intent recognition C o n d i t i o n [G) takes h-ld in the s a m e way t h a t it does for o t h e r i l l o c u t i o n a r y acts. n a m e l y in v i r t u e of a " f e a t u r e " of the u t t e r a n c e (e.g., u t t e r - ance mood, or a definite determiner} t h a t is c o r r e l a t e d with a c o m p l e x p r o p . s i t i o n a l a t t i t u d e . This a t t i t u d e b e c o m e s the basis for s u b s e q u e n t reasoning a b o u t the s p e a k e r ' s plans. In s u m m a r y , Searle's conditions can be a c c o n n t e d for by s i m p l y p o s i t i n g a n a c t i o n t h a t the s p e a k e r requests aml t h a t the h e a r e r reasons a b o u t a n d performs.

So far, the IAA a n d P A A are c o m p l e m e n t a r y . T h e y each a c c o u n t for different a s p e c t s of referring. T h e IAA characterizes u t t e r a n c e s who~e sole p . r p o s e is to secure referent identification, a n d the PAA c h a r a c t e r i z e s the use of referring p h r a s e s within a n i l l o c u t i o n a r y a c t . I now p r e c e d e to show how the IAA c a n s u b s u m e the PAA.

Searle a r g u e s t h a t tree use of the definite article in u t t e r i n g a n NP is to indicate the s p e a k e r ' s intention to refer uniquely. Moreover, from C o n d i t i o n 5, this intention is s u p p o s e d to be recognized by the hearer. We can get this effect by c o r r e l a t i n g the following expression with the delinite determiner:

A D [ ( M U T U A L - B E L I E F H e a r e r S p e a k e r ( G O A L S p e a k e r

( B E L H e a r e r

3 ! X ( G O A L S p e a k e r ( G O A L H e a r e r

( D O N E H e a r e r

IDENTIFY-REFERENT

(Hearer, D),

IDENTIFIED-REFERENT

(Hearer, D, X))))))]

T h i n k - f tiffs expression as being a p r a g m a t i c " f e a t u r e " of a syn- tactic c o n s t i t u e n t , as in c u r r e n t linguistic formalisms. W h e n this expression is applied to a d e s c r i p t o r (supplied from the semantics of the NP} we have a c o m p l e t e f o r m u l a t h a t b e c o m e s the seed for d e r i v i n g a request. Namely, if it is m u t u a l l y believed the s p e a k e r is sincere, 7 then it is m u t u a l l y believed there is a u n i q u e object t h a t s p e a k e r w a n t s the h e a r e r to w a n t to pick out. If it is m u t u a l l y believed the h e a r e r can do it {i.e.. the prec~mditions to the referent identification a c t hold, a n d the h e a r e r k n . w s how to do it by d e c o m p o s i n g the description into a plan c~f action), it is m u t u a l l y believed of s o m e object t h a t the s p e a k e r ' s goal is t h a t the h e a r e r a c t u a l l y pick it out. llence, a request. * T h u s . for the p e r c e p t u a l ease, the IAA subsumes the P A A .

5.1.

E x t e n d i n g t h e A n a l y s i s

A s s u m e t h a t instead of j u s t considering the a c t of identification in its p e r c e p t u a l sense, we a d o p t Searle's c o n c e p t - - n a m e l y t h a t *... there should no longer be a n y d o u b t w h a t e x a c t l y is being talked a b o u t . " Identification in this sense is p r i m a r i l y a process of establish- ing a coreferential link between the description in question a n d some o t h e r whose referent is in some way k n o w n to the hearer. However, we again r e g a r d identification as a n a c t t h a t the h e a r e r performs, not s o m e t h i n g the s p e a k e r does t o / f o r a hearer. If a n a n a l y s i s of this

"rSincerity can be dispensed with at no significant loss of generality

SThat is, I am suggesting that the interpretation of how the speaker intends the noun phrase to be interpreted le.g., referentially, attributively, etc.} begins with such a propositional attitude. If the referential reading is unsuccessful, the hearer needs to make other inferences to derive the intended reading.

e x t e n d e d notion c a n b e m a d e s i m i l a r in f o r m to the a n a l y s i s of the p e r c e p t u a l identification a c t , then the IAA c o m p l e t e l y s u b s u m e s the P A A . B e c a u s e both a c c o u n t s a r e equally v a g u e on w h a t c o n s t i t u t e s identification (as are, for t h a t m a t t e r , all o t h e r a c c o u n t s of which [ a m a w a r e ) , the choice between t h e m must rest on o t h e r g r o u n d s . T h e g r o u n d s favoring the identification r e q u e s t a n a l y s i s include the use of s e p a r a t e u t t e r a n c e s a n d i l l o c u t i o n a r y a c t s for referring, a n d the inde- p e n d e n t l y m o t i v a t e d s a t i s f a c t i o n of Searle's c o n d i t i o n s on referring.

5.2.

S e a r l e vs. R u s s e l l

Using the p r o p o s i t i o n a l a c t of referring. Searle a r g u e s a g a i n s t Rus- sews [13] t h e o r y of descriptions, which holds t h a t the u t t e r i n g of an expression " t h e ~" is equivalent to the assertion of an uniquely existential p r o p o s i t i o n ~there is a unique ~b". T h u s , when reference fails, it is b e c a u s e the uniquely existential p r o p o s i t i o n is nc, t true. Searle c l a i m s i n s t e a d t h a t the existence of the referent is a precondition to the a c t i o n of referring. In referring to X, we d o not a~ssert t h a t X exists a n y m o r e t h a n we d o in h i t t i n g X (lbid, p. 1G0.) However, the pre- c o n d i t i o n is n e c e s s a r y for successful p e r f o r m a n c e . Searle's a r g u m e n t a g a i n s t this t h e o r y essentially c o m e s down to:

. . . It [Russell's theory] presents the p r o p o s i t i o n a l a c t of definite reference, when p e r f o r m e d with definite descriptions . . . as equivalent to the i l l o c u t i o n a r y a c t of assert- ing a uniquely existential proposition, a n d there is no c o h e r e n t way to i n t e g r a t e such a t h e o r y into a t h e o r y of i l l o c u t i o n a r y acts. U n d e r no condition is a prol.J:~itional a c t identical with the i l l o c n t i o n a r y a c t of a s s e r t i . n , for a. p r o p o s i t i o n a l a c t c a n only o c c u r as p a r t of s, uuc illo- c u t i o n a r y a c t , never simply by itself (Ibid, p. 15. I

T h e r e a r e two difficulties with this a r g u m e n t . First, the require- " ment t h a t a c t s of referring be p a r t of a n ilk, c u t i o n a r y a c t was shown to be unnecessarily restrictive. Second, there is a way to assimilate the assertion of an existential p r o p o s i t i o n - - a n a c t t h a t .%arle claims does not c o n t a i n a referring a c t - - into an a n a l y s i s of illocutionary acts, n a m e l y as a n indirect r e q , e s t fur referent identification. How- ever, b e c a u s e a n assertion of a uniquely existential proposition may fail to c o n v e y an indirect request for referent identification (just as u t t e r i n g ~It's cold in here" m a y fail to c o n v e y an iuditert request} Searle's a r g u m e n t , t h o u g h weakened, still s t a n d s .

6. S u m m a r y

There are a number of advantages for treatitq~- refero:lt hh'ntifi-

c a t i o n as a n a c t i o n t h a t s p e a k e r s request, a n d thus f~Jr t r e a t i n g the

speech a c t of referring as a request. T h e analysis n - t ,,nly itrc,~tllltS for d a t a t h a t Searle's a c c o u n t can]tot, but it al.<o predicts each of Searle's c o n d i t i o n s for p e r f o r m i n g tlw a c t of s l n g n l a r idei~tifying reference, yet it allows for a p p r o p r i a t e e x t e u s h m into a Idaunitlg Impress. If we e x t e n d the p e r c e p t u a l use of referent identification t,~ Searl,.':~ more general c o n c e p t of identification, a n d we c o r r e l a t e a c e r t a i n {(;ricean) p r o p o s i t i o n a l a t t i t u d e with the use of definite d e t e r m i n e r s in a noun p h r a s e , then Searle's a n a l y s i s is s u b s u m e d by the a c t of requesting referent identification. T h e p r o p o s i t i o n a l a c t of referring is therefore unnecessary.

(5)

Ortony I10] has argued.

To keep this promise, we need to be specific about speaker- intentions for other uses of noun phrases. This will be no easy task. One difficulty will be to capture the distinction between achieving effects on a hearer, and doing so communicatively {i.e., in the Gricean way). Thus, for example, a hearer cannot comply with the illocutionary force of =Quick, don't think of an elephant* because there seems to be an "automatic" process of ~concept activation * 11]. Achieving ef- fects non-communicatively, i.e., without the recognition of intent, may be central to some kinds of reference. In such cases, speakers would be able to identify referents for a hearer. If this held for singular identifying reference, then tlwre could be grounds for a propositional act. llowever, we might have to give up the Gricean condition (5), which l suspect Searle would not want to do.

Finally, there are obviously many a.~pects of reference that need to be accounted for by any comprehensive theory. I make no claims {yet) about the utility of the present approach for dealing with them. Rather, 1 hope to have opened the door to a formal pragmatics for one aspect of referring.

7. R e f e r e n c e s

1. Appelt, D. Planning natural language utterances to satisfy mul- tiple 9oals. Ph.D. Th., Stanford University, Stanford, Califor- nia, December 1981.

2. Austin, J. L. How to do things with words. Oxford University Press, London, 1962.

3. Cohen, P. R. Pragmatics, speaker-reference, and the modality of communication. To appear in Computational Linguistics,

1984.

4. Cohen, P. R., & Levesque, If. J. Speech Acts and the Recog- nition of Shared Plans. Prec. of the Third Biennial Confer- ence, Canadian Society for Computational Studies of Intelli- gence, Victoria, B. C., May, 1980, 263-271.

5. Cohen, P. R., & Levesque, 11. J. Speech acts as summaries of shared plans. In preparation.

6. Donnellan, K. Reference and definite description. The Philo- sophical Re~Jieu, 7S. 1960, pp. 281-304.

7. Grice, 11. P. Meaning. Philosophical Review 66, 1957, pp. 377-

388.

8. ]lintikka, J. Semantics for propositional attitudes. In Philo- sophical lo#ic, D. Reidel Publishing Co., Dordrecht-Holland,1969.

9. Moore, R. C. Reasoning about knowledge and action. Techni- cal Note 191, Artificial Intelligence Center, SRI International, October, 1980.

10. Ortony, A. Some psycholinguistic constraints on the construc- tion and interpretation of definite descriptions. Proceedings of the Second Conference on Theoretical Issues in Natural Lan- guage Processing, Urbana, Illinois, 1978, 73-78.

I 1. Perrault, C. R., & Allen, J. F. A plan-based analysis of indirect speech acts. American Journal of Computational Linguistics 6,

3, 1980, pp. 167-182.

12. Perrault, C. R., & Cohen, P. R. It's for your own good: A note on inaccurate reference. In Elements of Discourss Understand- ing, Cambridge Irniversity Press, Joshi, A., Sag, I., & Webber, B., {Eds. , Cambridge, Mass., 1981.)

13. Russell, B. On denoting. Mind I4, 1905, pp. 479-492.

14. Searle, J. R. Speech acts: An essay in the philosophy of lan- guage. Cambridge University Press, Cambridge, 1969.

15. Sidner, C. L. The pragmatics of non-anaphoric noun plzrases. In Research in Knowledge Representation for Natural Language Understandin#: Annual Report, 9/1/8£-8/81/88, Bolt Beranek and Newman, Inc., , Cambridge, Mass, 1983.