Performatives in a Rationally Based Speech Act Theory*

Philip R. Cohen Artificial Intelligence Center and Center for the Study of Language and Information SRI International 333 Ravenswood Ave. Menlo Park, CA 94025 and Hector J. Levesque $ Department of Computer Science University of Toronto

Abstract 1 Introduction

There is something special about performative A crucially important adequacy test of any the- sentences, sentences such as "I promise to return": ory of speech acts is its ability to handle perfor- uttering them makes them true. How and when matives. This paper provides a theory of perfor- is this possible? Not all verbs can be uttered in matives as a test case for our rationally based the- the first-person present tense and thereby make ory of illocutionary acts. We show why "I request the sentence true. In general, the successful verbs you..." is a request, and "I lie to you that p" is seem to correspond to those naming illocution- self-defeating. The analysis supports and extends ary acts, but not to perlocutionary ones such as earlier work of theorists such as Bach and Harnish "frighten." But, even some illocutionary verbs [1] and takes issue with recent claims by Searle [10] cannot be used performatively: e.g., "I lie to you that such performative-as-declarative analyses are that I didn't steal your watch" is self-defeating doomed to failure. [12]. So, which verbs can be use performatively, and in Searle's words [10], "how do performatives *This paper was made possible by a contract from work?" ATR International to SRI International, by a gift from Any theory of illocutionary acts needs to pro- the Systems Development Foundation, and by a grant vide a solution to questions such as these. But, from the Natural Sciences and Engineering Research such questions are not merely of theoretical in- Council of Canada. The views and conclusions con- terest. Natural language database question- tained in this document axe those of the authors and answering systems have been known to receive should not be interpreted as representative of the of- ficial policies, either expressed or implied, of ATR In- performative utterances [14], dialogue systems ternational, the Systems Development Foundation, or that recognize illocutionary acts (e.g., [6]) will the Canadian government. need to infer the correct illocutionary force to t Fellow of the Canadian Institute for Advanced function properly, dialogue translation systems [5] Research. will have to cope with markers of illocutionary

79 force that function performatively (e.g., sentence tempt to base an analysis of performatives on such final particles in Japanese), and proposals for a theory of assertions. "agent-oriented programming languages" [7, 13], This paper provides a formal analysis of per- as well as Winograd and Flores' [15] COORDINA- formatives that treats them as declarative utter- TOR system, are based on performative communi- ances, not initially as assertions, does not succumb cation. For all these systems, it is important to to Searle's criticisms, and does not require an en- understand the semantics and pragmatics of such tirely new class of illocutionary acts (the "dec- communicative acts, especially their intended ef- larations") as Searle and Vanderveken [12] have fects. To do so, one needs a full theory of il- proposed. The analysis is offered as another ade- locutionary acts, and a formal theory that pre- quacy criterion for our theory of illocutionary acts. dicts how utterances can be made true by uttering That theory, more fully explicated in [3], is based them. on an analysis of the individual rational balance The currently accepted theory of performatives agents maintain among their beliefs, goals, inten- is that they are in fact assertions, hence true or tions, commitments, and actions [2]. false, and additionally constitute the performance of the named illocutionary act, in the same way As desiderata for the theory of performatives, as an indirect reading of an illocutionary act is we demonstrate that the analysis meets two prop- obtained from the direct illocutionary act. That erties: is, the named illocutionary act is derived from the assertion as an indirect speech act. The most com- • A sincere utterance of "I request you to open pelling defense of this performative-as-assertion the door" is both a request and an assertion, analysis that we are aware is that of Bach and Har- yet neither illocutionary act characterization nish [1], who address many of the linguistic phe- is derived from the other. nomena discussed by Sadock [9], but who, we believe, have misanalyzed indirect speech acts. How- ever, in a recent paper, Searle [10] forcefully crit- • "I lie that the door is open" is self-defeating. icizes the performative-as-assertion approach on the following grounds: Briefly, the ability to capture performatives is • Assertions commit the speaker to the truth met almost entirely because such utterances are of what is asserted treated as indicative mood utterances, and because illocutionary acts are defined as attempts. • Performative statements are self-referential Since attempts depend on the speaker's beliefs and • "An essential feature of any illocutionary act goals, and these mental states are introspectable is the intention to perform that act" in our theory if a speaker sincerely says, for example, "I request you to open the door," he must Searle claims that accounts based on self- believe he did the act with the requisite beliefs and referential assertions are "doomed to failure" be- goals. Hence, the utterance is a request. cause one cannot show that being committed to To meet the desiderata we need first to present, having the intention to be performing the named albeit briefly, the theory of rational interaction, illocutionary act entails that one in fact has that the treatment of declarative mood utterances, and intention. Moreover, he questions that one should then the illocutionary act definitions for request- derive the named illocutionary act from an asser- ing and asserting. Finally, we combine the vari- tion, rather than vice-versa. However, Searle has ous analyses natural language processor's task by imparted into Bach and Harnish's theory his no- making explicit the intended word sense of the action of assertions as commitments to the truth tion, and by reducing the combinatorics inherent without providing a precise analysis of commit- in determining the attachment of the prepositional ment. What may be doomed to failure is any at- phrases.

80 2 Abbreviated theory of rational Op def= -~<>-=p. action (PRIOR p q) dej Vc (HAPPENS c;q?) D Below, we give an abbreviated description of the 3a (a < c) A (HAPPENS a;p?). theory of rational action upon which we erect a The proposition p will become true no later than theory of intention. The theory is cast in a modal q. logic of belief, goal, action, and time. Further de- (KNOW x p) def= p A (BEL x p). tails of this logic can be found in [2]. (IMPERATIVE s) means that sentence s is an imperative. 2.1 Syntax (DECLARATIVE s) means that sentence s, a string The language we use has the usual connectives of a of words, is a declarative. first-order language with equality, as well as opera- tors for the propositional attitudes and for talking (MAIN-VERB s v), (TENSE s tense), (COMPLE- about sequences of events: (BEL x p) and (GOAL x MENT s s'), (D-OBJECT s np), (SUBJECT s np), p) say that p follows from x's beliefs or goals (a.k.a are all syntactic predicates intended to have the choices) respectively; (AGT x e) says that x is the obvious meanings. 2 only agent for the sequence of events e; el _reference, def and the meanings of imperatives, respectively. Their <~p -- =le (HAPPENS e;p?). last event arguments would be used only in the inter- (LATER p) d~f= ~p A Op. pretation of indexica]s. 4See other work of ours [2] for a full model theory. 1This differs from the BEFORE relation described 5In other words, we only deal with semantic struc- in [3], which is here labelled PRIOR. tures where these propositions come out true.

81 This means that agents choose amongst worlds Intending to do an action a or achieve a proposi- that are compatible with their beliefs. tion p is a special kind of commitment (i.e., per- Goals are known: sistent goal) to having done the action a or having I:::(GOAL x p) - (BEL x (GOAL x p)). achieved p.¢ However, it is not a simple commit- Memory: ment to having done a or e;p? for that would al- p (DONE x (BEL x p)?;e) = low the agent to be committed to doing something (BEE x (DONE x (BEE x p)?;e)). accidentally or unknowingly. Instead, we require That is, agents remember what their beliefs were. that the agent be committed to arriving at a state in which he believes he is about to do the intended action next. 3 Individual Commitments and In- This completes a brief discussion of the founda- tentions tional theory of intention and commitment. Next, we proceed to define the more specific concepts To capture one grade of commitment that an agent needed for analyzing communicative action. might have toward his goals, we define a persistent goal, P-GOAL, to be one that the agent will not give up until he thinks certain conditions are sat- 4 Utterance Events isfied. Specifically, we have We begin the analysis of utterance events by adopting a Gricean correlation of an utterance's Definition 1 (P-GOAL x p q) def= features (e.g., syntactic mood or sentence-final (1) (BEt x -~p) ^ particles in Japanese) with the speaker's mental (2) (GOAL x (LATER p)) A state, termed a "core attitude" in [3, 8]. Very (3) [KNOW x roughly, a declarative utterance $ will be corre- (PRIOR [(BEL x p)V(BEL x n-~p)v(eEL x "-,q)] lated with the speaker's believing the uttered sen- -~[GOAL x (LATER p)])]. tence is true, and an imperative utterance will be correlated with the speaker's wanting the ad- That is, the agent x believes p is currently false, dressee to do some action that fulfills the condi- chooses that it be true later, and knows that before tions imposed by the sentence. Let us notate these abandoning that choice, he must either believe it correlations as: is true, believe it never will be true, or believe q, DECLARATIVE =~ (aLL x (TRUE s e)) an escape clause (used to model subgoals, reasons, etc.) is false. IMPERATIVE =~ (GOAL x Intention is a species of persistent goal. We 03# (DONE y e') A analyze two kinds of intentions, those to do ac- (FULFILL-CONDS s e' e) tions and those to achieve propositions. Accord- We formalize this notation below. ingly, we define INTEND1 and INTEND2 to take Someone who thinks he is observing an utter- action expressions and propositions as arguments, ance event will come to believe the speaker is in respectively. the correlated mental state, unless he has other beliefs to the contrary. For example, if the ob- Definition 2 Intention: server thinks the speaker is lying, he believes that (INTEND1 x a q) def= the speaker does not believe the uttered sentence (P-GOAL x [DONE x (BEL x (HAPPENS a))?;a] q). is true. But, because he may think the speaker takes himself to be especially convincing, the ob- (INTEND~ x p q) def= server may still believe that the speaker thinks the (P-GOAL x observer is deceived. Hence, he would believe the 3e[HAPPENS x (BEE x 3e' (HAPPENS x e';p?))?;e;p?] 6For simplicity, we omit here one condition from q) the definition of INTEND2 in [2].

82 speaker thinks that he thinks the speaker believes state as a kind of default axiom (cf. [8]). First, we p. specify who is speaking (spkr), who is observing This type of reasoning can continue to further (obs, which includes the speaker and addressee, levels. In general, if an utterance is produced but also others), who is being addressed (addr), when there are no countervailing observer beliefs and what kind of sentence (s) has been spoken at a certain level of nesting, then the result will (indicated by q~). We shall assume that everyone be, at the given level of nesting, that the speaker knows that a given utterance is of a given syn- is taken to be in the correlated mental state [8]. tactic type (e.g., declarative), that speakers and To be able to state such conditions, we need to addressees are observers, and that observers are be able to refer easily to what a person x believes known by all to be observing. 7 about what y believes about what x believes etc., to arbitrary depths. To do so, we use the notion Definition 6 ~ =~ ~ de_/ of ABEL. V spkr, obs, addr, e, s, n (KNOW obs (DONE spkr e) A Definition 3 (ABEL n x y p) de__f (UTTER spkr addr s e) A (q~ s)) ^ (BEL x (BEL y (BEL x ...(BEL x p )...) ,-,(ABEL nobs spkr (BEFORE • That is, ABEL characterizes the nth alternating ,-,(GOAL spkr belief between x and y that p, built up "from out- [AFTER • side in," i.e, starting with x's belief that p. On (KNOW addr this basis, one can define unilateral mutual belief (BEFORE • o~))]) )) 2) -- what one agent believes is mutually believed -- (ABEL nobs spkr as follows. (BEFORE • t~ A (GOAL spkr Definition 4 (BMB x y p) def=Vn(ABEL n x y p) [AFTER • (KNOW addr In other words, (BMB x y p) is the infinite conjunc- (BEFORE • a))]) )) tion (BEL x p) A (BEL x (BEL y p)) ^... Finally, we define mutual belief and mutual knowledge as That is, • =~ ~ is an abbreviation for a quan- follows. tified implication roughly to the effect that if an Definition 5 (MB x y p) dej (BMB x y p) A observer obs knows that • was just done, where (BMB y x p). • was an uttering to addressee addt of a sentence s in syntactic mood q~, and obt does not believe (MKxyp) de---fpA(MBxyp). that • was done when the speaker did not want the addressee to come to know that the core speaker- Utterance events can produce effects at any (or attitude a associated with utterances of that type no) level of alternating belief. For example, the held, then obs believes that the speaker in fact speaker may not be trying to communicate any- wanted the addressee to know that o~, and so he, thing to an intended observer. Illocutionary acts the observer, believes that c~ held just prior to will be defined to require that the speaker intend the utterance. The notation states that at each to produce BM Bs. In what follows, it is important level of alternating belief for which the antecedent to keep in mind the distinction between utterance holds, so does the consequent. The symbol '=~' events and full-blooded communicative acts. can now be understood as a textual-replacement "macro" operator. 4.1 Notation for Describing Utterance Since these correlations are of the form Events VnP(n) 2~ Q(n)), they imply VnP(n) D VnQ(n). We now provide a formal notation for this corre- 7The case of unseen observers is straightforward, lation of utterance form and the speaker's mental but omitted here.

83 As we quantify over the positive integers indicat- fined in terms of the speaker's mental states. At- ing levels of alternative belief, we can derive the tempts involve both types of goal states, GOAL conclusion that under certain circumstances, addr (merely chosen) and INTEND (chosen with com- thinks it is mutually believed (in our notation, mitment), as noted below. BMB'ed) that the speaker spkr wants addr to know de] was true. Definition 7 {ATTEMPT x e p q tl} = Notice that right after the utterance, we are tI?;[(BEL x -,~p A ,,-q) A concerned with what mental state the observer (INTEND1 x tl?;e;p? (GOAL x Oq)) A thinks the speaker chose to bring about in the ob- (GOAL x Oq)]?; • server with that utterance. That is, the condition on utterance events involves the speaker's wanting to get the observer to know something, Without That is, an attempt to achieve q via p is a complex this temporal dimension, our performative analy- action expression in which x is the agent of event • sis would fail. The analysis of performatives will at time tl, and prior to e the agent believes p and say that after having uttered such a sentence, or q are both false, chooses that q should eventually while uttering it, the speaker believes he has just be true, and intends, relative to that choice, that • done or is doing the named illocutionary act. Typ- should produce p. So, q represents some ultimate ically, prior to uttering a performative, the speaker goal that may or may not be achieved by the at- has not just performed that speech act, and so he tempt, while p represents what it takes to make would believe his having just done so is false. So, if an honest effort. the condition on utterance events in Domain :Ax- iom 1A involved only what the speaker believed 5.1 Definitions of Request and Assert or wanted to be true prior to the utterance, rather than after, all performatives would fail to achieve To characterize a request or, for that matter, any the observer's coming to believe anything. illocutionary action, we must decide on the appro- We can now state the correlation between ut- priate formulas to substitute for p and q in the def- terance form and a speaker's mental state as a inition of an attempt. We constrain illocutionary domain axiom. acts to be those in which the speaker is committed to understanding, that is, to achieving a state of Domain Axiom 1 Declaratives and BMB that he is in a certain mental state. Below is Imperatives: a definition of a speaker's requesting an addressee to achieve p. A. ~=DECLARATIVE =~ (BEL spkr (TRUE s e)) Definition 8 {REQUEST spkr addr • p tl} def= B. I= IMPERATIVE :=~ (GOAL x O3e'(DONE y e') ^ {ATTEMPT spkr • (FULFILL-CONDS s e' e) [BMB addr spkr (BEFORE • (GOAL spkr Below, we present our definitions of illocutionary Op A acts. Further justification can be found in [3]. [AFTER • (INTEND~ addr p 5 Illocutionary Acts as Attempts [(GOAL spkr Op) A (HELPFUL addr spkr)] )])]] Searle [11] points out that an essential condition 3e' (DONE adclr e';p?) for requesting is that the speaker be attempting to tl} get the addressee to perform the requested action. We take this observation one step further and de- That is, event • is a request at time tl if it is fine all illocutionary acts as attempts, hence de- an attempt at that time to get the addressee to

84 achieve some condition p while being committed exactly as declarative utterances, with the excep- to making public that the speaker wanted: first, tion that the content of the utterance will make that p eventually be achieved; and second, that reference to an utterance event. the addressed party should intend to achieve it relative to the speaker's wanting it achieved and 6.1 Request Reports relative to the addressee's being helpfully disposed towards the speaker. Let us characterize the truth conditions of the The illocutionary act of asserting will be defined family of declarative sentences "x requests y to as an attempt to make the speaker's believing the (imperative sentence sl). " Let s be such a sen- propositional content mutually believed. tence. Let ct be 3el(DONE y el) A (FULFILL- CONDS s' ez e). We ignore most syntactic con- def siderations and indexicality for reasons of space. Definition 9 {ASSERT spkr addr • p tl} = {ATTEMPT spkr addr • Domain Axiom 2 Present tense requests [BMB addr spkr J= Vx, y, e, tl, (DONE h?;e) ^ (BEFORE e (SUBJECT s ~) A (D-OBJECT s y) A [GOAL spkr (REFERS z x e) A (REFERS y y e) D (AFTER • (TRUE s e) - (DONE x {REQUEST x y e ~ tl}) [KNOW addr (BEFORE • (BEL spkr p))])])] That is, if event • is happening and the sentence s (BMB acldr spkr (BEFORE e (BEL spkr p))) is a present tense declarative sentence whose main h} verb is "request," whose subject x refers to person x, and whose direct object Y refers to person More precisely, assertions at time tl are defined y, then the sentence is true iff x is requesting the as attempts in which to make an "honest effort," addressee y to fulfill the conditions of imperative the speaker is committed to getting the addressee sentence s'. A bare present (or present progres- to believe that it is mutually believed that the sive) tense sentence is true when the event being speaker wanted prior to the utterance that the described is contemporaneous with the event of addressee would come to know that the speaker uttering it. s This definition applies equally well believed p held then. That is, just like a request, to "John requests Mary to ..." as it does when I an assertion makes public that the speaker wants utter "I request you to ..." For the former, such the addressee to know what mental state he was sentences are likely to be narrations of ongoing in. Although he is committed to that, what the events. 9 For the latter, the event that is happen- speaker has chosen to achieve is not merely to ing that makes the utterance true is the speaker's make public his goal that the addressee know what uttering of the sentence. mental state he was in, but to make public that By our definition of request, for x to request he was in fact in that state of believing p. For y to achieve p, x has to attempt to get y to do an INFORM, the speaker would choose to achieve some action intentionally to fulfill the sentence s', (KNOW addr p). by making that goal mutually believed between them. Thus, to say x requested y to do something is only to say that x had the right beliefs, goals, 6 Performatives and intentions. To illustrate how performatives work, we show SSearle [10] correctly points out that performatives when both assertions and requests can be derived can be uttered in the passive, and in the first-person from the utterance of the performative "I request plural. you to ." The important point to notice 9We are ignoring the habitual reading of bare here is that we have not had to add to our ma- present tense sentences because we do not have a se- chinery; performative utterances will be treated mantics for them.

85 6.2 Performatives Used as Requests 6.3 Performatives Used as Assertions Next, we treat performative sentences as declar- We have shown that the speaker of a sincere per- atives. This means that the effects of uttering formative utterance containing an illocutionary them are described by Domain Axiom 1A. We verb has performed the illocutionary act named by sketch below a proof of a general theorem re- that verb. Under somewhat stronger conditions, garding performative requests, with s being the we can also prove that the speaker has made an declarative sentence"I request you to(imperative assertion. As before, let s be "I request you to sentence Sl) , and c~ being 3el(DONE addr el) A ." (FULFILL-CONDS S1 e I e). We take the uttering of a sentence to be a unitary utterance event. Theorem 2 Perforrnatives Used as Assertions I::V spkr, addr, e, n, tl, Theorem 1 A Performative Request (MK spkr addr (DONE spkr tl?;e) A I=V spkr, addr, e, n, tl, (UTTER spkr addr • s)) A (MK spkr addr (DONE spkr tl?;e) A (UTTER spkr addr e s)) A [BEFORE • (BEFORE h?;e (BEL spkr (GOAL spkr [AFTER e [AFTER tl?;e Vn~,(ABEL n addr spkr (KNOW addr (BEFORE e [BEFORE tl?;e ~(GOAL spkr (BEL spkr (TRUE s e))])])) Z) [AFTER • (KNOW addr (DONE {REQUESTspkr addr e a tl}) [BEFORE • That is, we need to show that if the sentence "I (BEL spkr (TRUE s e))]] request you to " has just been uttered at time tl sincerely, i.e., when the This default condition says that before the ut- speaker wanted the addressee to know that he be- terance, the speaker believed there would be no lieved the sentence was true, then a direct request addressee belief after the utterance event (at any has taken place at tl. level n) to the effect that prior to that event the Proof sketch: Essentially, one invokes the do- speaker did not want the addressee to come to main axiom for declaratives at the first level of know that the speaker believed (TRUE s e). Given ABEL, entailing that the speaker believes that he Domain Axiom 1A, and the fact that BEL entails believes that he has just done a REQUEST. Then, GOAL, this suffices to entail the definition of asser- one expands the definition of REQUEST into an tion. Notice that whereas requesting was derived ATTEMPT, and then into its parts. The defini- in virtue of the content of the utterance, an assertion of ATTEMPT is based on BEL, GOAL and tion was derived by default assumptions regarding INTEND, the first two of which are obviously in- lack of belief in the speaker's insincerity. trospectable. That is, if one believes one has them one does, and vice-versa. Hence, by the memory 7 'Lie' is not a performative assumption, the speaker actually had them prior to the utterance. More critically, intending to act Some illocutionary verbs such as "lie, .... hint, .... in- at time tl is also introspectable at time tl because sinuate," cannot be achieved performatively. The agents know what they are doing at the next in- following analysis shows a general model for why stant and because there is no time to drop their such verbs naming covert acts cannot be perfor- commitment [2]. Thus, one can repackage these matively achieved. mental states up into an ATTEMPT and then a A reasonable definition of lying is the following REQUEST. complex action:

86 Definition 10 {LIE spkr addr e p} de__f GOAl', and INTEND the addressee can conclude (BEL spkr ~p)?;{ASSERT spkr addr e p tl} that, before the utterance, the speaker wanted him not to know that the speaker believes that in uttering S, he was lying. But, this contradicts the That is, a lie is an assertion performed when the assumption (**). Since the speaker in fact uttered speaker believes the propositional content is false. the sentence, that assumption is false, and the ad- For "I lie to you that the door is open" to be dressee believes the speaker did not in fact want a successful performative utterance, it would have him to know that he believed the sentence was to be true that the utterance is a lie. We would true. This renders impossible the intentions to be have to show that the uttering of that declarative achieved in asserting, which are constitutive of ly- sentence results in a lie's having been done. More ing as well. generally, we provide a putative statement of the Now, assume (**) is false, so the addressee in truth conditions of "x lies to y that ." Call the main sentence s. that s' was true. Again, this immediately makes the speaker's intentions in asserting, and hence ly- Domain Axiom 3 Supposed Truth Conditions ing, impossible to achieve. So, in neither case is for Performative Lying the utterance a lie. If the addressee believes the l:: Ve, x, y, tl, (DONE h?:e) A (REFERS x x e) A speaker is a competent speaker of the language, (REFERS y y e) D the speaker must have intended some other inter- (TRUE s e) - pretation. (DONE {LIE x y e (TRUE s' e) tl} )

That is, if s and s' are declarative sentences of 8 Conclusion the appropriate syntactic form, x refers to x and Requesting works well as a performative verb be- y refers to y, then s is true iff in performing it at cause requesting requires only that the agent has time tl, x was lying that sentence s' is true. made an attempt, and need not have succeeded in So we can prove the following. Let the sentence getting the hearer to do the requested action, or s be "I lie to you that ." even to form the right beliefs. Some verbs cannot be used performatively, such as "frighten," Theorem 3 Lies are not performative ~V spkr, addr, e, n because they require something beyond a mere (MK spkr addr [(DONE spkr tl?;e) A attempt. Hence, such verbs would name action (UTTER spkr addr • s)]) D expressions that required a particular proposition ,-,(DONE {LIE spkr addr e (TRUE s e) tl}) p be true after the utterance event. When the utterance event does not guarantee such a p, the use In other words, you cannot perform a lie by saying of the performative verb will not be possible. "I lie that ..." On the other hand, certain utterances (perfor- Proof Sketch: Assume that it is mutually be- mative or not), when performed by the right peo- lieved that the speaker has uttered declarative ple in the right circumstances, make certain insti- sentence s. Now, apply Domain Axiom 1A. By tutional facts hold. So, when a clergyman, judge, assumption, the first conjuct of the antecddent or ship captain says "I now pronounce you hus- holds. There are then two cases to consider. First, band and wife," the man and woman in question assume (**) the second conjunct holds (say, at are married. In our framework, there would be a level n = 1), i.e., the addressee does not believe domain axiom whose antecedent characterizes the the speaker did not want him to know that he be- circumstances, participants, and nature of the ut- lieved s' was true. In virtue of the supposed truth terance event, and whose consequent asserts that conditions on lying, spkr would have to have been an institutional fact is true. The axiom is justified lying. By expanding its definition, and using the not by the nature of rational action, but by the ex- memory and introspectability properties of BEI_, istence of an institution. Such utterances could be

87 made with a performative prefix provided such at- [7] J. McCarthy. ELEPHANT: a programming tempts are made into successes by the institution. language based on speech acts. Unpublished This paper has shown that treating performa- ms., Dept. of Computer Science, Stanford tive utterances as declarative sentences is a vi- University, 1989. able analysis, in spite of Searle's criticisms. The [8] C. R. Perrault. An application of default logic performative use of an illocutionary verb is self- to speech act theory. In P. R. Cohen, J. Mor- guaranteeing when the named illocutionary act gan, and M. E. Pollack, editors, Intentions consists in the speaker's making an attempt to in Communication. M.I.T. Press, Cambridge, make public his mental state. In such cases, if Massachusetts, in press. the speaker thinks he has done so, then he has. However, we do not derive the named illocution- [9] J. Sadock. Toward a Linguistic Theory of ary act from the assertion, nor vice-versa. Instead, Speech Acts. Academic Press, New York, both derivations may be made from the utterance 1984. event, but the assertive one is in fact harder to [10] J. Searle. How performatives work. Linguis- obtain as it has extra conditions that need to be tics and Philosophy, 12:535-558, 1989. satisfied. [11] J. R. Searle. Speech acts: An essay in the philosophy of language. Cambridge Univer- References sity Press, Cambridge, 1969. [1] K. Bach and R. Harnish. Linguistic Com- [12] J. R. Searle and D. Vanderveken. Founda- munication and Speech Acts. M. I. T. Press, tions of lllocutionary Logic. Cambridge Univ. Cambridge, Massachusetts, 1979. Press, New York City, New York, 1985.

[2] P. R. Cohen and H. J. Levesque. Intention [13] Y. Shoham. Agent oriented programming. is choice with commitment. Artificial Intelli- Unpublished ms., Dept. of Computer Science, gence, 42(3), 1990. Stanford University, October 1989. [14] H. Tennant. Evaluation of natural language [3] P. R. Cohen and H. J. Levesque. Rational processors. Technical Report T-103, Coordi- interaction as the basis for communication. In nated Science Laboratory, University of Illi- P. R. Cohen, J. Morgan, and M. E. Pollack, nois, Urbana, Illinois, November 1980. Ph. editors, Intentions in Communication. M.I.T. D. Thesis. Press, Cambridge, Massachusetts, in press. [15] T. Winograd and F. Flores. Understanding [4] D. Harel. First-Order Dynamic Logic. Computers and Cognition: A New Founda- Springer-Verlag, New York City, New York, tion for Design. Ablex Publishing Co., Nor- 1979. wood, New Jersey, 1986. [5] K. Kogure, H. Iida, K. Yoshimoto, H. Maeda, M. Kume, and S. Kato. A method of analyzing Japanese speech act types. In Second International Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages, 1986.

[6] D.J. Litman and J. F. Allen. A plan recogni- tion model for subdialogues in conversation. Technical report, Department of Computer Science, Univ. of Rochester, Rochester, New York, November 1984.