Meaning and speech acts

©

Keith Allan

Linguistics Program, Monash University

If we adopt illocutionary point as the basic notion on which to classify uses of language, then there are a rather limited number of basic things we do with language; we tell people how things are, we try to get them to do things, we commit ourselves to doing things, we express our feelings and attitudes, and we bring about changes through our utterances. Often we do more than one of these at once in the same utterance.    (Searle 1975a:369)

1. WHERE WE ARE HEADING

Today, our interest in speech acts stems directly from the work of Austin who came from the Oxford school of "ordinary language philosophers" which also spawned Geach, Ryle, Strawson, Grice, and Searle. It was intellectually engendered by Wittgenstein, who observed (e.g. 1953 §23) that, to that time, logicians had had very little or nothing to say about many of the structures and usages in natural language. The analysis of speech acts is an account of contextualized utterance meaning, cf. Austin 1962¹ , Searle 1969, 1979, Labov and Fanshel 1977, Bach and Harnish 1979, Edmondson 1981, Recanati 1987, Allan 1994b. Most important is the illocutionary point of an utterance, i.e. what Speaker intends to achieve by making the particular utterance U_i in the particular context C_i. We need identify the process for linking the senses of the language expressions used to the meaning of Speaker's utterance: we begin by clarifying what is meant by Speaker and Hearer, and defining reflexive intention; then we identify the hierarchy of acts that create the messages in utterances. To this end, 'Meaning and speech acts' consists of the following sections:

 2.    Speakers, Hearers and overhearers
 3.    Locutions, illocutions, and perlocutions within the hierarchy of acts
 4.    Performative clauses
 5.    Speech act classification and definition
 6.    Presuppositions and preconditions on illocutions
 7.    Clause-type and primary illocution
 8.    The inference schema for calculating illocutionary point
 9.    On-record and off-record, literal and nonliteral
10.    Cultural diversity
11.    More than one illocution in an utterance
12.    Speech acts and discourse
13.    Summary
14.    References

2. SPEAKERS, HEARERS AND OVERHEARERS

The social-interactionist view is that language results from acts of speaking or writing when someone (whom we shall refer to as Speaker) says (or writes) something to someone else (whom we shall refer to as Hearer) at a certain time in a certain place -- often as part of a longer discourse or interchange. We shall refer to these acts of speaking and writing as SPEECH ACTS (Austin 1962, Searle 1969, 1979, Bach and Harnish 1979, Allan 1994e). There is an assumption that Speaker intends to communicate with Hearer. As recognized by Grice 1957, 1968, 1969 the intention is reflexive: it is Speaker's intention to have a person in earshot recognize that Speaker wants him or her to accept the role of Hearer and therefore be an (or the) intended recipient of Speaker's message and consequently react to it.

Definition. Speaker's REFLEXIVE INTENTION toward Hearer is the intention to have Hearer recognize that when uttering U in context C, Speaker intends U to have a certain effect on Hearer partly caused by Hearer recognizing that Speaker has the intention to communicate with him or her by means of U (cf. Grice opp.cit. and Recanati 1987).

So, when Joe hears Sue talking in her sleep, he will not assume she has a reflexive intention towards him, and therefore not expect that she intends her utterance to have any effect on him -- though she might unintentionally keep him awake. There are innumerable mental, emotional, and physical effects that speakers might wish to produce, e.g. persuading Hearer to an opinion, intimidating Hearer, alerting Hearer of danger, getting Hearer to do something by means of a suggestion, a hint, a request, or a command.

In the spoken medium there is never more than one speaker per utterance; however, two speakers may utter identical utterances in unison, or Speaker may speak on someone else's behalf. Co-authors generally take joint responsibility for what is written but, normally, each writes only a part of the text. This starkly contrasts with the number of hearers or readers Speaker may have for an audience.

Definition. Hearer is anyone whom, at the time of utterance, Speaker reflexively intends should recognize the illocutionary point of the utterance (= Speaker's message).

Speaker tailors the utterance to suit Hearer, taking into account the presumed common ground and what s/he knows or guesses about Hearer's ability to understand the message s/he wants to convey.

Clark and Carlson 1982 distinguish between Hearer as 'direct addressee', and Hearer as 'ratified participant', the latter being a member of the audience participating in the speech act (cf. Goffman 1981:131). The notion of face (Brown and Levinson 1987) is useful in distinguishing between two kinds each of hearers and overhearers. An ADDRESSEE is someone who cannot reject the role of Hearer without serious affront to Speaker. Direct address is determined contextually -- by direction of gaze, pointing a finger, touching an arm, using a name, or on the basis of who spoke last; less commonly, the nature of the message will determine who is the intended addressee. Note the change of addressee in 1 and the nonspecific addressee in 2.

1. Joan, Max bought me this beautiful ring for our anniversary, didn't you Max, you sweetie!
2. Congratulations, whoever came first!

A RATIFIED PARTICIPANT can reject the Hearer role more freely than an addressee and with less of an affront to Speaker. When Speaker is speaking, all those who can reasonably consider themselves ratified participants are expected, as part of the cooperative endeavour, to keep tabs on what is said, so that if called upon to participate they may do so appropriately.

Any other person hearing U is an OVERHEARER: either a bystander or an eavesdropper. People in earshot are expected to overhear, though not necessarily to listen; only hearers are properly expected to listen. It can happen that U is overheard by someone when there was no original specific intention on Speaker's part that this should happen; to put it more precisely, Speaker has a reflexive intention towards Hearer but not towards an overhearer. An overhearer may perchance understand the message the same way Hearer does because they share common ground; but, because s/he is not necessarily party to the appropriate contextual information relevant to the correct interpretation of the utterance, it is possible that s/he may seriously misinterpret it. So, a BYSTANDER within earshot was not originally intended as a Hearer and may, depending on circumstances, accept or reject the role of Hearer without loss of face; consider an occasion where X is arguing with Y in earshot of Z.

An EAVESDROPPER can only admit to listening at the risk of looking bad, and perhaps also affronting Speaker.

To top

3. LOCUTIONS, ILLOCUTIONS, AND PERLOCUTIONS WITHIN THE HIERARCHY OF ACTS

Communication is successful not when hearers recognize the linguistic meaning of the utterance, but when they infer the speaker's "meaning" from it. (Sperber and Wilson 1986:23)

Making a normal utterance involves a hierarchy of acts. To begin with, there is the ACT OF UTTERANCE. We recognize utterance acts, even in a language that is completely unknown to us, in which we cannot distinguish the sentences used, and what Speaker's message is. We do this on the basis of brute perception: by hearing the utterance spoken, seeing it signed or written, or feeling it impressed in braille. Linguistics is concerned with utterances in which Speaker uses a language expression and thereby performs a locutionary act (and more).

Definition. In performing a LOCUTIONARY ACT Speaker uses an identifiable expression e, consisting of a sentence or sentence fragment from language L, spoken with identifiable prosody, p. We symbolize the form of the locution <π, e> and say that it has the sense "μ".²

Prosody π is composed of the pattern of pause, pitch level, stress, tone of voice, and the like; its counterpart in the written medium is punctuation and typography. μ is the intension of <π, e>; normally, a locution demands that Speaker and Hearer have knowledge of the grammar, lexicon, semantics, and phonology of L.

Definition. Speaker's ACT OF REFERRING applies locution <π,e>, meaning "μ" to a particular world w_i at time t_i.

Austin, whose How To Do Things With Words (1962) first awakened wide interest in speech acts, included the act of referring as part of the locutionary act (1975:109), and they were first separated by Searle in Speech Acts (1969:81ff). Whereas locutions are defined on a particular language, reference is defined on particular worlds. Different Speakers using different locutionary and utterance acts can refer to the same thing; for instance, at a gathering in which there are speakers of English (4), Swahili (5), and Tohono O'odham (6), 4-6 could all be referring to the same dog.

4. The dog's barking!
5. Mbwa anabweka!
6. Hi:nk o g gogs!

Obviously, under normal conditions of use, Speaker makes an utterance, uses a locution, and refers with it, all at one and the same moment.

The most significant act in the hierarchy of speaking is the illocutionary act. Austin 1962 alerted us to the fact that Speaker DOES something in uttering U to Hearer in context C, e.g. states a fact or an opinion (Semantics can be difficult), confirms or denies something (It's not true that Marilyn Monroe committed suicide), makes a prediction (It'll rain tonight), a promise (I'll be with you in five minutes), a request (What's the time?), offers thanks or an invitation (Can you come to dinner next Saturday?), issues an order or an umpire's decision (Out!), gives advice or permission (Yes, of course you can leave early today), names a child or a ship (I name this ship "QE3"), swears an oath (I swear allegiance to the King).

Definition. In utterance U, Speaker performs an ILLOCUTIONARY ACT in using a particular locution to refer, such that U has the ILLOCUTIONARY FORCE of a statement, a confirmation, a denial, a prediction, a promise, a request, and so forth.

We shall see later that an utterance has more than one illocutionary force; but many utterances have only one message to convey, and this particular illocutionary force is the ILLOCUTIONARY POINT. In 7, the locution is what you see following the example number.

7. I'll make the tea.

The person who utters it and the context of utterance will determine the reference. One illocutionary force is that of a statement about a future act. In a given utterance of 7, it may be used with a second illocutionary force: to make a promise. If this is the recognized intention of Speaker, then that promise is the illocutionary point of the utterance. (There is a more detailed account of this process later.)

The illocutionary point of 8 would typically be to have Hearer recognize that Speaker is offering a bet; the acceptance or refusal of the challenge is the PERLOCUTIONARY EFFECT of the utterance.

8. I bet you a dollar you can jump that puddle.

Definition. Speaker's PERLOCUTIONARY ACT is act of achieving a particular perlocutionary effect on Hearer as a result of Hearer recognizing (what s/he takes to be) the locution and illocutionary forces in U.

So, a perlocution is Hearer's behavioural response to the meaning of U -- not necessarily a physical or verbal response, perhaps merely a mental or emotional response of some kind. Other perlocutions are such things as: alerting Hearer by warning Hearer of danger; persuading Hearer to an opinion by stating supporting facts; intimidating Hearer by threatening; getting Hearer to do something by means of a suggestion, a hint, a request, or a command; and so forth. An effect of U which does NOT result from Hearer recognizing the locution and illocutionary point of U is NOT a perlocutionary effect, but some kind of gestural effect (e.g. responding to a raised voice or an angry look). Perlocutions are extremely significant within a theory of communication because the normal reason for speaking is to cause an effect in Hearer, and Speaker typically strives to achieve this by any means s/he can. However, perlocutionary effects fall beyond the boundary of linguistics because they are not part of language but behavioural and/or cognitive and/or emotional responses to the illocutions in utterances. What linguists can properly look at, however, are the intentions of speakers to bring about certain perlocutionary effects:

Definition. Speaker's intention to cause a perlocutionary effect on Hearer by having Hearer recognize the illocutionary forces in Speaker's utterance is variously referred to as Speaker's PERLOCUTIONARY INTENTION (Bach and Harnish 1979) or Speaker's ILLOCUTIONARY INTENTION (§5 below and Allan 1986, 1994e).

To top

4. PERFORMATIVE CLAUSES

The constative utterance, under the name so dear to philosophers, of statement, has the property of being true or false. The performative utterance, by contrast, can never be either: it has its own special job, it is used to perform an action. To issue such an utterance is to perform the action -- an action, perhaps, which one scarcely could perform, at least with so much precision, in any other way. Here are some examples:

I name this ship 'Liberté'.
I apologise.
I welcome you.
I advise you to do it.       (Austin 1963:22)

Austin's point is that in making such utterances under the right conditions, Speaker performs, respectively, an act of naming, and act of apologizing, an act of welcoming, and an act of advising (today it is usual to speak of 'acts' rather than 'actions').

Necessary Condition 1. An explicit performative clause complies with the normal grammatical rules of the language and contains a verb that names the illocutionary point of the utterance.

In 9, Speaker uses an explicit performative clause (italicized) to make a promise. Speaker could also have made the promise by uttering 10, in which the promise is not explicitly spelled out in the semantics of the verb but is inferred by means we examine later.

9.  I promise to call Jo tomorrow.
10. I'll call Jo tomorrow.

Here is a short list of performative verb listemes (there are many more): abjure, abolish, accept, acknowledge, acquit, admit, admonish, advise, affirm, agree to, announce, answer, apologize, ascribe, ask, assent, assert, assess, assume, authorize, baptize, beg, bet, bid, call upon, caution, charge, christen, claim, classify, command, commiserate, compliment, concur, congratulate, conjecture, convict, counsel, declare, declare out, delegate, demand, demur, deny, describe, diagnose, disagree, dispute, donate, dub, excuse, exempt, fire, forbid, give notice, grant, guarantee, guess, hire, hypothesize, identify, implore, inform, instruct, license, name, notify, offer, order, pardon, permit, plead, pray, predict, prohibit, promise, proscribe, query, question, rank, recommend, refuse, reject, renounce, report, request, require, rescind, resign, sanction, say, sentence, state, submit, suggest, summon, suppose, swear, tell, testify, thank, urge, volunteer, vouch for, warn, withdraw.

Necessary Condition 2. The performative verb must be in the present (nonpast, nonfuture, nonperfect) tense, because the illocutionary act is defined on the moment of utterance.³

Contrast performative 11 with the same listeme used nonperformatively in 12.

11. I promise to take Max to a movie tomorrow.
12. I promised to take Max to a movie tomorrow.

Saying 'I promise' in 11, Speaker makes a promise; but the words 'I promised' in 12 do not constitute the making of a promise; instead, they report that a promise was made.

Necessary Condition 3. (In English) a performative may occur in either the simple or progressive aspect.

A performative verb normally occurs in the simple aspect, perhaps for the same reason that the simple aspect is normal in on-the-spot reporting of football matches, baseball games, etc. However, there are occasions where a performative may occur in the progressive aspect:

13. I am requesting you (for the umpteenth time) to tell me your decision.
14. I request you (for the umpteenth time) to tell me your decision.

13 has the illocutionary point of a request: the grounds for claiming it to be a statement about a request are no stronger than the grounds for claiming the same about 14. Uttered in appropriate circumstances (see below on 'felicity conditions'), 15 has the illocutionary point of a bet, so Hearer can justifiably reply You're on! thereby taking up the bet, and expecting Speaker to pay up when s/he loses -- or vice versa.

15. That horse has won its third race in a row, and I'm betting you $100 it'll win on Saturday.

Necessary Condition 4. A performative clause must be declarative and realis "real, actual, factual" i.e. denote an actualization of the illocutionary act.

An explicit performative clause cannot be interrogative, imperative, or subjunctive. None of 16-18 is performative:

16. Shall I bet $50 on the cup?
17. Get out of here!
18. Should I recommend her for the job?

Condition 4 also places constraints on the modal auxiliaries that may occur in performative clauses and so there is a link with tense. Past and present tenses locate things that have happened or are happening because past and present times are realis. However, the future is IRREALIS "unreal, unactualized, nonfactual", and this links the future with subjunctive mood and modality (cf. Lyons 1977, Bybee, Perkins and Pagliuca 1994, Bybee and Fleischman (eds) 1995b). Thus, in English, the typical indicator for the future is the modal auxiliary will, cf. Fig.1.

Figure 1. The time-tense-reality network

19. I will hereby promise to visit you next time I'm in town.

In 19 the modal will is used in its 'root' meaning "act on one's will, desire, want, hence insist" on carrying out the illocutionary act named in the performative verb promise. 19 denotes an ongoing act that can be glossed "I will with these words make the promise to visit you next time I am in town"; so if Max utters 19 to his aged aunt but then doesn't visit next time he is in town, his aunt can justifiably chide him with breaking his promise: But you promised to visit! Contrast the performative promise of 19 with the predicted promise in 20.

20. Tomorrow when I see her, I will promise to visit next time I'm in town.

Sufficient Condition. The legalistic-sounding adverb hereby, inserted into a performative clause, will mark the verb as performative provided that hereby is used with the meaning "in uttering this performative" (cf. Austin 1962).

Hereby cannot legitimately be inserted between 'will' and 'promise' in 20; which confirms that 'promise' is not a performative verb. The modal will is used in its epistemic "predict" sense⁴ and is irrealis because it denotes an unactualized event, namely the future act of promising (to take place 'tomorrow').

The pattern established by will holds generally for modal auxiliaries with performative verbs which actualize the illocutionary act. The modal must therefore be used in its root meaning, which is realis, cf. the leave-taking in 21 and the warning in 22. 23 is ambiguous:

21. I must hereby take my leave of you.
22. Trespassers should hereby be warned that they will be prosecuted.
23. I can hereby authorize you to act as our agent from this moment.

The root meaning of can and could is linked to the adjective cunning and north British dialect canny: "actor knows how and has the power and ability, to do act A". In 23, if 'can' means "have the power to" and 'hereby' means "in uttering this performative", then 23 effects an authorization ("I have the power by the utterance of these words to authorize you ..."). However, if 'I can hereby' means "using this fax from head-office makes it possible for me to", then 23 is not an authorization but a statement about a possible authorization. 24-26 are additional examples of nonperformative clauses with modals.

24. I might promise to visit you next time I'm in town.
25. I might hereby authorize your release.
26. I could hereby sentence you to ten years imprisonment.

Modal might is never realis and it is obvious that 24 states the possibility that Speaker will promise without actualizing a promise. The 'hereby' that occurs in 25 necessarily has the sense "using this", and refers to something in context other than the performative utterance, e.g. a confession from another party; thus 25 is nonperformative. Similarly 26 does not pass sentence; compare it with I hereby sentence you to ten years imprisonment. In 26 'could' is epistemic and irrealis, and 'hereby' once again means "using this".

Necessary Condition 5. The subject of the performative clause is conditioned by the fact that Speaker is agent for either him- or herself or another person or institution, whichever takes responsibility for enforcing the illocution described by the performative verb.

This influences the form of the subject noun phrase. In all examples so far, the subject was I. However, 27-28 have 'we'; 29-30 are passive voice, and the authorization is made either on Speaker's behalf or behalf of someone else; there is a third person subject in 31 where an authorized agent utters the performative on behalf of the court.

27. We, the undersigned, promise to pay the balance of the amount within ten days.
28. We hereby authorize you to pay on our behalf a sum not exceeding $500.
29. You are hereby authorized to pay ...
30. Notice is hereby given that trespassers will be prosecuted.
31. The court permits you to stand down.

In 31 'permits' is performative because it is the issuing of this utterance which actually grants the permission.

Explicit performatives can be negative. For instance the illocutionary force of a refusal can be borne by either I refuse your request or, less likely, by the negative performative in 32, where 'I don't grant' is uttered at a relatively higher pitch than 'your request'.

32. I don't grant your request.

Contrast this with the same sequence of words uttered with a disjuncture after 'don't', and a lower pitch level for 'grant your request' (33) which is a paraphrase of 34. 33-34 are not refusals, but statements about a refusal.

33. I don't / grant your request.
34. I don't [as you claim] 'grant your request'.

35 performs an act of not-promising; note the scope of the negative: an act of not-promising is different from an act of promising not to do something as in 36.

35. I don't promise to come to your party, but I'll try to make it.
36. I promise not to come to your party.

Austin 1962, 1963 insisted on a distinction between what he called CONSTATIVES, which have truth values, and performatives which don't, cf. the quote at the head of this section. The distinction between truth-bearing and non-truth-bearing sentences has a long history. Aristotle noted that 'Not all sentences are statements [apophantikos]; only such as have in them either truth or falsity. Thus a prayer is a sentence, but neither true nor false' (On Interpretation 17a,1). Later, the Stoics distinguished a 'judgment' (axioma) as either true or false whereas none of an interrogation, inquiry, imperative, adjurative, optative, hypothetical, nor vocative has a truth value (cf. Diogenes Laertius Lives VII:65-68). For more than two millennia, logicians and language philosophers concentrated their minds on statements and the valid inferences to be drawn from them, to the virtual exclusion of other propositional types (questions, commands, etc.) Austin was reacting to this tradition (cf. Hare 1971 Ch.6). Austin believed that instead of truth values, performatives have FELICITY CONDITIONS: in his opinion, 37 has no truth value but is felicitous if there is a cat such that Speaker has the ability and intention to put it out, and is infelicitous -- but not false -- otherwise. This contrasts with 38 which is either true if Speaker has put the cat out, or false if not.

37. I promise to put the cat out.
38. I've put the cat out.

Austin's claim that performatives do not have truth values has been challenged from the start (cf. Cohen 1964, Lewis 1970, Bach 1975) and he seems to have been misled by the fact that their truth value is less communicatively significant than their illocutionary point.

Austin (1975:14f) argued for four kinds of felicity conditions:

a.       A PREPARATORY CONDITION to establish whether or not the circumstances of the speech act and the participants in it are appropriate to its being performed successfully.

b.       An EXECUTIVE CONDITION to determine whether or not the speech act has been properly executed.

c.       A SINCERITY CONDITION -- which has a similar function to the cooperative maxim of quality.

d.       A FULFILLMENT CONDITION determined by the perlocutionary effect of the speech act.

If all the relevant felicity conditions were satisfied for a given illocutionary act, Austin described it as 'happy' or 'felicitous'. We can immediately dismiss (d) as beyond the boundary of a linguistic theory of speech acts because it has only a contingent link with the language used. Preparatory conditions, (a), are exemplified in §5 and discussed at length in §6 under the abbreviated name PRECONDITIONS. As for executive conditions, (b), Austin requires that the procedure invoked by the illocutionary act 'must be executed by all participants correctly and completely' (1975:14). He exemplifies this with I bet you the race won't be run today said when more than one race was arranged for that day. But such misexecutions should be dealt with under generally applicable maxims of the cooperative principle. Today, the only executive condition needed is for a class of speech acts known as 'declarations' (see §5) which bring about, or express decisions upon, states of affairs such as marriage, job appointment and termination, and umpiring. These rely for their success on Speaker being sanctioned by a community or institution to perform the acts under conditions stipulated in the preparatory conditions on the act: they identify the attitude or behaviour that must be observed by Speaker when executing the illocutionary act in order for it to be felicitous. An executive condition safeguards society's interests by seeking to ensure that the sanctions are respected.

(c), the sincerity condition on a speech act, involves Speaker's responsibility for the illocutions in the utterance. If Speaker is observing the cooperative maxim of quality, then s/he will be sincere: and, normally, Hearer will assume that Speaker is being sincere unless s/he has good reason to believe otherwise. Generally, scholars have assumed that different kinds of illocutionary acts involve different kinds of sincerity conditions: e.g. assertions and the like are sincere if Speaker believes in the truth of the proposition asserted; requests are sincere if Speaker believes Hearer has the capability and might be willing to carry out the act requested; declarations are sincere if Speaker believes s/he has the proper authority to make the declaration (cf. Austin 1962, Searle 1969, Bach and Harnish 1979, Allan 1986). Sincerity reflects on whether or not Speaker upholds the preconditions, so only one sincerity condition is necessary:

Definition. The GENERALIZED SINCERITY CONDITION: in uttering U, Speaker knows or believes (or believes s/he knows) that all clauses of the precondition hold.

This puts a burden on precise statement of the preconditions; but that seems exactly where it should lie, because preconditions identify the particular circumstances appropriate to performing a given illocutionary act (§§5-6).

To sum up: the only one of Austin's original felicity conditions that remains obligatory in the definitions of all illocutions (see §5) is the precondition, though an executive condition may be valid for declarations. Sincerity can be captured by generally applicable conditions on language use. Finally, linguists rarely attend to fulfillment conditions, nor should they -- though these will remain important in disciplines concerned with perlocutionary effects of utterances. The burden of felicitous illocution will depend on proper observation of the preconditions on each illocutionary act. These conditions provide the grounds for motivating Speaker to make the utterance, and grounds from which Hearer will evaluate the illocutionary act expressed in the utterance.

To top

5. SPEECH ACT CLASSIFICATION AND DEFINITION

There have been two approaches to classifying speech acts: one, following Austin, is principally a lexical classification of so-called illocutionary verbs; the other, following Searle 1975a, is principally a classification of acts. Austin 1962 identified five classes of illocutionary verbs which were refined and extended to seven by Vendler 1972 as follows. Glosses are from Austin 1975:151-61 and structural descriptions are from Vendler, whose N="NP", V="verb", p="proposition (sentence)", nom="gerund" or other nominalization.

Expositives 'expounding of views, the conducting of arguments and the clarifying of usages and of references' (N_i V that p), e.g. state, contend, insist, deny, remind, guess
Verdictives 'the giving of a verdict' (N_i V N_j (as) N or Adj) e.g. rank, grade, call, define, analyze.
Commissives 'commit the speaker' (N_i V to V) e.g. promise, guarantee, refuse, decline.
Exercitives 'exercising of powers, rights or influences' (N_i V N_j to V) e.g. order, request, beg, dare
Behabitives 'reaction to other people's behaviour and fortunes' (N_i V N_j P nom (past (V))) e.g. thank, congratulate, criticize.

Vendler's two extra classes:

Operatives (N_i V N_j to be/become N_k ) e.g. appoint, ordain, condemn.
Interrogatives (N_i V wh-nom (p)) e.g. ask, question.

Katz (1977:50-57) critically examines Vendler's analysis. A more extensive lexical classification is Ballmer and Brennenstuhl 1981: they present a thesaurus-like lexicon where verbs are grouped according to an illocutionary property such as Make a hidden appeal, which includes bitch at, carp about, grumble, murmur, mutiny, nag, pout, rumble, sulk, whine, and wrangle (p.73); and there is Put someone to flight, which includes chase away, chase off, discharge, dismiss, drive away, drive back, force out, frighten away, kick out, oust, put to flight, scare off, see off, send packing, squeeze out, and throw out (p.100). Ballmer and Brennenstuhl identified speech act verbs using the formula exemplified in 39-40.

39. Jo ______ "Why me?" [e.g. whined]
40. Jay ______ "Don't let me see you here again!" [e.g. chased him off with]

Their classification of 4,800 verbs into 600 categories comprising 24 types in eight groups was made (as it is by other scholars) on an intuitive basis in terms of semantic similarity. Wierzbicka 1987 makes a much more explicit semantic analysis of 270 speech act verbs, grouping them into 37 classes. E.g. her Promise class contains promise, pledge, vow, swear, vouch for, and guarantee.

Searle 1975a lists 12 differences between speech acts that can serve as bases for classification, but he uses only four of them to establish five classes of speech acts.

(a) ILLOCUTIONARY POINT (see the quote at the head of this paper). For instance, a request attempts to get Hearer to do something; an assertive is a representation of how something is; a promise is the undertaking of an obligation that Speaker do something.
(b) DIRECTION OF FIT between the words uttered and the world they relate to: e.g. statements have a words-to-world fit because truth value is assigned on the basis of whether or not the words describe things as they are in the world spoken of; requests have a world-to-words fit because the world must be changed to fulfill Speaker's request.
(c) THE EXPRESSED PSYCHOLOGICAL STATE: e.g. a statement that p expresses Speaker's belief that p; a promise expresses Speaker's intention to do something; a request expresses Speaker's desire that Hearer should do something.
(d) PROPOSITIONAL CONTENT: e.g. Hearer to do A (i.e. perform some act) for a request; Speaker to do A for a promise. The five kinds of speech act Searle recognizes are:

Assertives (called representatives in Searle 1975a) have a truth value, show words-to-world fit, and express Speaker's belief that p.
Directives are attempts to get Hearer to do something, therefore they show world-to-words fit, and express Speaker's wish or desire that Hearer do A.
Commissives commit Speaker to some future course of action, so they show world-to-words fit, and Speaker expresses the intention that Speaker do A.
Expressives express Speaker's attitude to a certain state of affairs specified (if at all) in the propositional content (e.g. the bolded portion of I apologize for stepping on your toe). There is no direction of fit; a variety of different psychological states; and propositional content must be related to Speaker or Hearer (1975a:357f).
Declarations bring about correspondence between the propositional content and the world; thus direction of fit is both words-to-world and world-to-words. Searle recognizes no psychological state for declarations.

Bach and Harnish (1979:42-51, 110f) employ all of Searle's criteria except direction of fit, giving prominence to Speaker's psychological state -- which they refer to as Speaker's 'attitude'. They identify six classes, splitting Searle's 'declarations' into effectives which effect changes in institutional states of affairs; and verdictives which have official binding import in the context of the institution for which they are made. Declarations (to use Searle's term) are different from the other classes of 'interpersonal' acts.

Definition. Assertives, directives, commissives, and expressives are INTERPERSONAL ACTS, typically directed at individuals. To take effect, they require Hearer to react to Speaker's illocution -- mere understanding of the illocutionary point is insufficient.

It is pointless for Speaker to tell Hearer it is raining, warn Hearer of danger, or offer Hearer condolences if Hearer fails to react appropriately to what Speaker says.

Definition. DECLARATIONS are typically broadcast within a social group, and rely for their success on Speaker being sanctioned by the community, institution, committee, or even a single person within the group to perform such acts under stipulated conditions. Provided the stipulated conditions are met, Hearer's reaction as an individual is irrelevant to the effectiveness of the declaration (e.g. being baptized, disqualified from driving, or fired).

Significant for declarations is the reaction of the group which sanctions Speaker. Compare the interpersonal opine that p (e.g. I think history is bunk) with the declaration declare the verdict p (e.g. Speaker, umpiring a game at the U.S. Open Tennis Tournament, declares the ball Out!)

Because declarations rely for their success on Speaker being sanctioned by the community, etc. it may be necessary to safeguard society's interest with an executive condition which requires some watchdog other than Speaker to ensure that clauses (ii)-(v) of the precondition hold. Note the legalistic character of the declaration.

Speech acts can be grouped into four classes if we use HEARER'S EVALUATIONS as criteria (cf. Allan 1994b,c).

Statements (including denials, reports, predictions, promises, and offers) can all be judged in terms of the question "Is p credible?" These are principally expressions of Speaker's belief about the way the world was, is, or will be, and are most typically formulated with a declarative clause.
Invitationals are a proper subset of Searle's directives, and include requests, exhortations, suggestions, warnings, etc. They have acceptability values: "Does Speaker really want A done, and if so is Hearer both able and willing to do it?" These principally invite Hearer's participation, and many are formulated in an interrogative clause.
Authoritatives include the rest of Searle's 'directives' and his 'declarations' (i.e. commands, permissions, legal judgments, baptisms, etc.) for which Hearer must consider the question "Does Speaker have the authority to utter U in this context?" These principally have Speaker "laying down the law"; many of them are formulated in an imperative clause, the rest in a declarative.
Expressives (greetings, thanks, apologies, congratulations, etc.) have social- interactive-appropriacy values: "Has something occurred which warrants Speaker expressing such a reaction to it?" These principally express social interaction with Hearer; many are idiomatic, the rest are in the default declarative clause format.

It is notable that all classes of speech acts can be conveyed using a declarative clause; but interrogatives typically indicate invitationals, imperatives typically indicate authoritatives, and idioms typically indicate expressives.

A comparison of these different classifications is Fig.2.

Figure 2. A comparison of five classifications of illocutionary types

Happily, there is great similarity in the speech act definitions of Searle 1969, Bach and Harnish 1979, Edmondson 1981, Levinson 1983, Allan 1986, and Wierzbicka 1987 despite the different perspectives. To identify the critical components of a definition we review six definitions of assertives. For easy comparison, clauses in the definitions which serve similar functions are identified by a common label.

Searle's 'propositional content' is the complement of the speech act verb and excludes its meaning -- unlike the 'description' ascribed to other scholars. Searle's term for 'illocutionary intention' is 'essential condition'.

By 'intention' they mean a "reflexive intention".

All these definitions are an expansion of the semantics of the key verb (e.g. assert) naming the illocution. However, most utterances do not contain such a verb, and Hearer must recognize the illocution without its help; the way this is accomplished is discussed later. Most scholars refer to utterance content, identified above as 'description' and 'propositional content'. Searle's definition refers to the content of the proposition that is asserted (p) and, unlike most other scholars, does not include a description of utterance content as a whole. He (correctly) believes that the same proposition may be used to express different illocutionary forces which are indicated outside of p by what he calls 'illocutionary force indicating devices' (IFIDs) captured in his 'essential condition'. This causes a problem, for instance, in his definition of thank (for), where the propositional content is 'Past act A done by H' (Searle 1969:67). Corresponding components in other definitions of thank are

'In uttering U, S thanks H for D if S expresses ...'    (Bach and Harnish 1979:52)
'S expresses thanks to H for D'    (Allan 1986)
'I say: I feel something good towards you because of that'    (Wierzbicka 1987:214)

We see that Searle's propositional content has nothing in particular to do with thanking. Moreover, it is inaccurate for utterances such as Thank you for joining me or Thanks. If you plan to adopt Searle's theory, you need to choose between rectifying his definition and abandoning his 'propositional content' component in favor of a <description> which will include it.

Both Searle's second precondition, 'It is not obvious to both S and H that H knows (does not need to be reminded of, etc.) p' and Wierzbicka's 'I imagine some people would say this is not true' offer reasons for Speaker making the assertion rather than keeping quiet (Wierzbicka identifies a class of people who disagree with Speaker's belief and whom Speaker hopes to persuade that p is true; this point is applicable to the semantics of assert but not of state). Instead of including such conditions in the fine detail of an individual semantic definition, they can be stated within a general set of cooperative presumptions applicable to all speech acts.

Searle and Wierzbicka identify a sincerity condition; this is not a necessary component of the definition, because sincerity can be subsumed to the cooperative maxim of quality applicable to every speech act in a form given in §4.

The illocutionary intentions for assertions fall into two parts: the first identifies Speaker's commitment to belief that p; the second, Speaker's presumed intention towards Hearer. The latter can be inferred from general cooperative presumptions about Speaker's purpose in uttering the assertion to Hearer, and so does not need stating in the definition.

We see from the review of six definitions of assertives that their obligatory components are description, precondition, illocutionary intention. The same holds true for all other classes of speech act; for definitions of which, the reader should turn to the works cited.

To top

6. PRESUPPOSITIONS AND PRECONDITIONS ON ILLOCUTIONS

VINCENTIO: Come hither, you rogue. What, have you forgot me?
BIONDELLO: Forgot you! No, sir. I could not forget you, for I never saw you before in all my life.

(Shakespeare Taming of the Shrew V.i.49)

The statement of what Austin and Searle call 'preparatory conditions', which we abbreviate to PRECONDITIONS, is obligatory in definitions of illocutions. The preconditions identify what ought to be presupposed in a felicitous use of the illocution, cf. Karttunen and Peters 1979:10. In this section we argue that presuppositions are preconditions on illocutions.

Everyone who knows a little of French history knows that 41 is false in the world we inhabit.

41. In 1990, France had a bald King.

France did not have a bald King in 1905 either, when Bertrand Russell claimed that 42 is false.

42. The present King of France is bald.

Both 41 and 42 entail that the King of France has extension at the relevant time.

41a. There was in 1990 a King of France.
42a. There is a King of France.

Both 41a-42a are false, hence, by modus tollens 41-42 are false. 42a is also entailed by all of 43(a-e), thus they too are false (or similarly faulty if such a valuation is not appropriate to 40(b-e)).

43.

(a) The present King of France is not bald.
(b) Where's the King of France gone?
(c) Is the King of France attending Di and Chuck's wedding?
(d) Shoot the King of France!
(e) That damned King of France!

In the tradition of Frege 1892, Strawson 1950 said that a sentence like 42 does not entail but falsely PRESUPPOSES⁵ 42a. The fact that both 42 and 43(a) presuppose 42a gives rise to the standard definition of presupposition, which uses the classical relation of entailment.

Definition of semantic presupposition. If A entails B and not-A entails B, then A presupposes B.

In addition to definite NPs, there are many other kinds of presupposition triggers: Karttunen identified 31 according to Levinson 1983:181 (cf. Gazdar 1979:127). We list just a couple here. Factive predicates such as know, realize, regret, be aware, be glad, be indifferent, be proud, that, be odd, be sad, be sorry presuppose the truth of their complements.

44.

Jo regrets (doesn't regret) kissing Eric presupposes Jo kissed Eric.
Does(n't) Jo regret kissing Eric? presupposes Jo kissed Eric.

Verbs of judging presuppose their sentential complements:

45.

Sue accused (didn't accuse) Ed of having an affair with Shiela presupposes Ed has been having an affair with Shiela
Sam blamed (didn't blame) Suzie for their being late presupposes Sam and Suzie were late
Mike forgave (didn't forgive) Jim for speaking out of turn presupposes Jim spoke out of turn

Implicative verbs like forget (cf. the quote at the head of this section), happen to, have the misfortune to, manage presuppose the truth of their complements in declarative clauses, but not in interrogatives and imperatives. Did Jack manage to speak to Harry does not presuppose Jack spoke to Harry though it does presuppose There is a question about whether or not Jack spoke to Harry. Aspectual verbs like begin, continue, stop, come to and iteratives like return, repeat, another time have aspectual presuppositions.

46.

Harry started (didn't start) to eat presupposes Harry had not been eating before
Jerry stopped (didn't stop) seeing Elaine presupposes Jerry had been seeing Elaine
Harry repeated (didn't repeat) the year presupposes Harry had done the year before
Jack rang (didn't ring) for a third time presupposes Jack had already rung twice

Cleft and heavily stressed NPs have presuppositions such as in 47:

47.

It was(n't) Jack who called presupposes Someone called
JACK went (didn't go) presupposes Someone went

There is logical problem with the standard definition of presupposition:

Problem. By the standard definition, A presupposes B = either A or else not-A entails B

In bivalent propositional logic, 'either A or else not-A' is necessarily true in every logically possible world; it is a tautology and truth conditionally equivalent to every other tautology.

(a)    A tautology cannot (without contradiction) entail a nontautology, therefore what is presupposed must be a tautology.
(b)    Every sentence entails every tautology.
(c)    It follows that every sentence presupposes all tautologies (cf. Katz 1973:258, Boër and Lycan 1976:6).

This violates the intuitive notion of presupposition, and despite the fact that the definition of semantic presupposition appears to account for many cases of presupposition, its consequences are unacceptable.

For a variety of technical reasons Gazdar 1979 and Levinson 1983 argue that semantic presupposition is simply not viable. One, already suggested in Strawson 1950, is that pragmatic presupposition is essentially speaker based in the sense that we interpret A presupposes B as "A indicates that Speaker presupposes B". But it is not always Speaker who does the presupposing; for instance with proprioceptive verbs of thinking and believing, or with verbs of saying:

48. Sam says (thinks, believes, writes) that the King of France is coming to his party, which just goes to prove the guy is nuts.

It is 'Sam' who presupposes there's a King of France; the 'which...' clause implicates that Speaker doesn't. Karttunen 1973 named verbs of this kind 'presuppositional PLUGS'.

Presuppositions are pragmatic because they can be cancelled, unlike semantic/logical relations. For example, 49 cancels the presupposition of 42 and 50 the presupposition in 44.

49. The present King of France isn't bald, because there is no King of France at present.
50. Jo doesn't regret kissing Eric, because in fact she never kissed him in the first place.

These negations have the peculiar property of being refutations that are intuitively quotative -- because it is appropriate to insert words to the effect of as you claim before the comma in 49-50. Many people have distinguished two types of negation: an ordinary negative which does not negate the presupposition, e.g. Jo doesn't regret kissing Eric, and what Seuren 1985 calls a RADICAL NEGATIVE (also misleadingly called 'external' negation e.g. in Gamut 1991) which denies the presupposition as well as the assertion (cf. 49-50). In radical negation the presupposing clause has to be negated, retention of an affirmative creates incoherence as in 51-52.

51. *The present King of France is bald, because there is no King of France at present.
52. *Jo regrets kissing Eric, because in fact she never kissed him in the first place.

The difference between 49-50 and 51-52 arises from the fact that the because clauses in 51-52 entail the falsity of the presupposition in the first clause creating a self- contradiction similar to the phenomenon known as Moore's paradox (Moore 1952:542f), cf. 53 where 'this' = "linguistics is less rigorous than philosophy".

53.

*Linguistics is less rigorous than philosophy, but I don't believe this is true.
*Linguistics is less rigorous than philosophy and I don't know this for a fact!

54.

I don't believe that linguistics is less rigorous than philosophy.
I don't know for a fact that linguistics is less rigorous than philosophy.

In order to observe the cooperative maxim of quality and sincerely utter Linguistics is less rigorous than philosophy, Speaker must (at least pretend to) believe that the proposition is true: all 51-53 are contradictory because Speaker specifically denies this necessary precondition on his or her felicitously uttering the sentence (i.e. Speaker can cancel an apparent presupposition but not deny a presupposition s/he implicitly holds to be true).

Imagine the stress, intonation and tone of voice in which you would utter any of 55.

55.

The door wasn't green [as Harry claimed], it was blue.
George isn't tall and thin, he's short and fat; you're thinking of the wrong man!
There is no King of France, because France is a republic.

All of them are explicitly contrastive and quotative. It is the quotative ('metalinguistic') characteristic of radical negation that distinguishes it from ordinary negation (cf. Seuren 1985:238ff, Horn 1989), but this is not sufficient reason to identify a formal contrast between them. Furthermore, NO language has one morphological form for regular negation and a different form for radical negation. We must therefore conclude that there is only one kind of negation, but we shall retain the phrase radical negation to describe the negating of presuppositions.

The difference between 49 and 51 is proof enough that, when recognized, failed presupposition is indicated by negation. A purportedly true assertion such as The present King of France is bald contradicts an accompanying recognition of its presupposition failure; consistency requires it to be marked as not-true. Strawson 1950 suggested that statements with false presuppositions have no truth value. Quine 1960 dubbed this a 'truth-value gap' which gives rise to trivalent truth {T, F, #} where # symbolizes "neither true nor false". There are undoubtedly sentences whose truth value is difficult to determine; not only disputed sentences like 41-43 naming a nonexistent King of France, but also those like I've got my mires wuddled and word salads such as This blue and on speak conferenced uply. While none of the foregoing examples are true, they are not indisputably false either. To recognize this fact, we follow Strawson and Quine to admit two kinds of not-true assertions:

Hypothesis. Those sentences that are not-true are either false or INDETERMINABLE.

Indeterminable sentences are best dealt with as a pragmatic problem. One kind of indeterminability arises from a failed precondition on the illocution. For instance, the precondition on France is a republic is "Speaker has reason to believe that France is a republic". If Speaker had said France is NOT a republic we would condemn Speaker for being ignorant, deluded, insane, or maliciously attempting to mislead Hearer. This is exactly the normal response to sentences such as 41-46 that exhibit presupposition failure. In such cases, Austin would say 'the utterance is void' (1975:51). Condemnation as a response to precondition failure is common to all illocutions. Take the precondition on thanking: "Hearer, or someone or something in Hearer's charge, has done some deed D with the apparent intention of benefiting Speaker (directly or indirectly)"; if Speaker thanks Hearer for D when Hearer never did D, Hearer will either conclude that Speaker is deluded or is being sarcastic. Finally, take the case of the tennis player in the U.S. Open who claims his/her opponent's ball is Out when the umpire disagrees: the claim has no standing because the player is not a person sanctioned by the preconditions on this illocutionary act to declare the ball out of play. Again, by violating the precondition, Speaker risks condemnation.

Some inferences that have been described in the literature as presuppositions are more properly to be regarded as either entailments or conventional implicatures because they cannot be cancelled. They do, however, (like all entailments and conventional implicatures) correspond to preconditions on Speaker's use of the illocution. Most propositions in temporal clauses are entailed (pace Levinson 1983:182).

56.

When Zeno was born, isn't known entails Zeno was born
Harry cooked the fish while Sally made the chips entails (Sally made the chips) & (Harry cooked the fish)
Sam told me about the accident before John called entails (John called) & (Sam told me about the accident)

However, the semantics of the main clause may modify the entailment (cf. Heinämäki 1972, Gazdar 1979).

57. Jack died before he finished writing the book entails (Jack died) & (Jack was writing a book)

Here Jack's death terminates an ongoing activity before it reaches its proper conclusion, cf. Jack hit the off-switch before the engine blew and The car broke down before they got to Wagga Wagga (cf. Gazdar 1979). Many CONVENTIONAL IMPLICATURES of verbs have been called presuppositions, e.g. by Fillmore 1971, Lakoff 1971, Levinson 1983. In 58-61, impossible cancellations are given in square brackets.

58. credit NP with/for X conventionally implicates X is praiseworthy & NP is responsible for X

James credited his co-author with the invention that made them millions [*but he didn't believe his co-author had done anything praiseworthy].

59. call NP a bitch conventionally implicates insult NP [denoting a female human, or a gay male human, or an inanimate object]

Max called the young lady a bitch and she bridled at the insult [*but he wasn't insulting her].

60. say NP looks terrific conventionally implicates compliment NP

Fred told Zelda she looked terrific, and she complimented him in return [*but in fact he was insulting her].

61. accuse NP of X conventionally implicates NP deliberately did reprehensible act X

Sue accused Ed of having an affair with Shiela [*but neither Speaker nor Sue thought there was anything wrong with it].

The conversational implicature of 61 is that both Speaker and Sue believe X is reprehensible; it can be cancelled for one, but not both of them. Blame NP for X is similar.

In this section we have rejected the view that presupposition is a well-defined semantic or logical property. The weight of evidence is that presuppositions are the preconditions on illocutions, and presupposition failure arises through a speaker's inability or failure to ensure the correct preconditions for what is said. Such failure causes Speaker to be judged ignorant, deluded, insane, or maliciously trying to mislead Hearer. Like conversational implicatures, presuppositions can be cancelled. If presuppositions arise from the preconditions on illocutions, then they are subject to the maxim of quality. For instance, someone who utters 42 implicates that s/he has reasonable grounds to believe that the proposition it contains is true and therefore implicates 42a. Similarly, the maxim of quality safeguards the implicature from say, 44, that Jo kissed Eric; and so forth for all the examples we have given of presupposition.

Hypothesis. Presuppositions are those conversational implicatures that:

(a) hold under negation, and
(b) relate specifically to the preconditions on the speech act.

The hypothesis that presuppositions are a kind of conversational implicature explains why they share in the feature of cancellability.

To top

7. CLAUSE-TYPE, MOOD, AND PRIMARY ILLOCUTION

The form of the utterance must be the starting point for Hearer's interpretation of its meaning. Thus, in order to fully understand an utterance U Hearer must recognize (a-c):

(a)    The locution <π,e> meaning "μ".
(b)    What Speaker intends the locution to refer to.
(c)    The illocutions within U.

Hearer hears the locution, recognizes what Sperber and Wilson in the quote above refer to as 'the linguistic meaning', look to the context to figure out the apparent reference, and then seek to infer Speaker's illocutionary intention (Sperber and Wilson's 'speaker's "meaning"'. Thus intial or primary illocution is signalled by the locution.

Suppose a herpetologist utters one of 62-65 to his or her spouse. 62-65 have a common propositional content: "Hearer's not ever handling the cobra"; the difference in meaning is indicated by the different clause-types or moods (S_n symbolizes "Speaker of n").

62. You never handle the cobra. [S₆₂ makes a statement using the declarative]
63. Do you never handle the cobra? [S₆₃ asks a question using the interrogative]
64. Never handle the cobra! [S₆₄ issues an imperative]
65. (a) Would that you never handle the cobra! (b) If only you were never to handle the cobra ... [S₆₅ expresses a wish using the subjunctive]

Making a statement, asking a question, issuing an imperative, and expressing a wish are distinct illocutionary acts; i.e. Speaker is doing something different in each of 62-65. Because the same proposition is used, the difference must be a function of the different mood in each sentence. Hence:

Hypothesis. MOOD (clause-type) is the initial clue to determining the illocutionary point of the utterance because mood indicates the primary illocutionary force of an utterance.

Thus mood has a very important role in speech act theory.

Grammarians in the western classical tradition have recognized a degree of coincidence between clause-type and illocutionary force at least since the time of Apollonius Dyscolus (100 CE, cf. Householder 1981:12f) and probably since 300 BCE (cf. Diogenes Laertius 'Life of Zeno' VII, 65-68; also Sanctius 1585, Lancelot 1644, Lane 1700, Whitney 1888, Sadock 1974:16f.). Palmer's 1986 monograph distinguishes the interrogative within the modality system, but Lyons 1977 argues against the identification of mood with clause-type because in the western classical tradition both the declarative and the interrogative are indicative in mood. However, there are not only interrogatives like 63 asking about the actual world, but also subjunctive interrogatives like 66 which ask questions about hypothetical worlds:

66. Would you never handle the cobra?

We shall return to the matter of the two interrogatives shortly.

We have identified mood with clause-type and in 62-66 demonstrated that mood, and therefore clause-type, indicates the illocutionary force. Fragments like declarative John in answer to Who's that?, interrogative John? [Is that you?], imperative JOHN! [Where the hell are you!], subjunctive If only! [that were so] are not a problem.

Sadock and Zwicky 1985 surveyed 35 languages representing a wide range of language families and linguistic areas: every one of them distinguishes a declarative to (among other things) make statements, an interrogative to ask things of people, and an imperative to get them to do things. These three moods are orthographically marked by '.', '?', and '!' respectively. Many languages have clause types with other functions -- e.g. optative- subjunctive, expressive-exclamative, prohibitive, imprecative. English has not only the (optative-)subjunctive in 65 and 66, but also expressive(-exclamative) in mostly monolexical nonverb idioms of two kinds: those not necessarily addressed to anyone, 67, and those addressed to Hearer, 68. Note that some imprecatives are listed.

67.

Goodness gracious!
Shit!
Wow!

68.

Thanks.
Please.
Sorry.
Pardon.
Hi.
Bye.
Congratulations.
Asshole!

Some of 68 have idiomatic counterparts with verbs:

69.

Thank you.
How do you do?
Screw you!

Imprecations of third persons appear to be expressives (70-71), but the clauses in 72 are imperative:

70.

X: Suzie found Tom in her own bed with another woman!
Y: Asshole! ["Tom is an asshole"]

71. Fuck him! [NOT= "You copulate with him"]

72.

Apologize to him for me.
Congratulate her.
Thank him.

Often the expressives of 67-71 have similarly idiomatic counterparts in other languages.
We represent mood as an operator on the propositional content of the utterance. For instance, the generalized form of the primary illocution of a DECLARATIVE clause is as follows:

73.

(a) Jack is bald.
(b) Frank loathes Harry.
(c) [Q: Who called? A:] Jack.

We refer to the primary illocution of the declarative clause as a 'declarative' which is not to be confused with a 'declaration'. The declarative performs an act of saying; a declaration performs acts of declaring, effecting, etc. The generalized form of the primary illocution of an INTERROGATIVE clause is:

74.

(a) Will you mail this?
(b) What's the time?

Bearing in mind that many imperative clauses are not at all imperious (cf. 75) the primary illocution of an IMPERATIVE clause is:

75.

(a) Forgive me intruding.
(b) Excuse me.
(c) Let me help you with that.
(d) Have a beer.
(e) Take the first turning on your left and the third on the right.
(f) Have a good day!

The primary illocution of an EXPRESSIVE is:

The primary illocution of a SUBJUNCTIVE is:

76.

(a) I wish I were rich.
(b) I wish I was rich.
(c) Would that I were rich.
(d) If Harry should call, tell him I'll be back this evening.

The subjunctive environment is the complement of a verb of wishing or wanting, invoking the hypothetical world. In 76(a) the past tense form were is also the subjunctive form, but the past tense itself is used to mark the subjunctive in (b) -- the tenses of all 15 are semantically nonpast. A past tense subjunctive would be had been. Note the use of modals would and should in (c) and (d). Conditional if marks a subjunctive clause when the proposition invokes a hypothetical world, so that should can be utilized in a paraphrase -- e.g. (d) is a paraphrase of If Harry calls, ....

English interrogative subjunctives are restricted to requests with only four backshifted modals, cf. 77.

77.

(a) Would you mail this for me?
(b) Could you do me a favour?
(c) Might he be there by now?
(d) Should I write to him?

They are all notably tentative which accounts for their use in polite contexts. They have the following properties:

The form of the interrogative subjunctive suggests that the 'indicative' interrogative has an interrogative operator ranging over a declarative.

The primary illocutions fall into four classes corresponding to Searle's 1975a notion 'direction of fit':

Declaratives show a words-to-world fit: the words match the way the world was or is or will be.
Requests and imperatives both show a world-to-words fit: things are to happen in the world to make it match the propositional content. They differ because Speaker expressly gives Hearer the option not to comply when using a request; whereas with imperatives, Speaker expressly gives Hearer no option but to comply.
For expressives, direction of fit is irrelevant.
Subjunctives fit the words to a hypothetical world.
Interrogative subjunctives seek to fit the actual world to a hypothetical world: things are to happen in the actual world to make it match the words that describe the hypothetical world.

The recognition of mood identifies the primary (or initial) illocution in the utterance, but not Speaker's illocutionary point. It is often assumed that performative clauses, in which the verb spells out the illocutionary point, express their illocutionary point directly (e.g. Gazdar 1981); but the analysis of 78 makes this impossible: the primary illocution of a performative clause is that of a declarative (Cohen 1964, Lewis 1970, Bach and Harnish 1979, Allan 1986).

78. I promise to go there tomorrow.

In 78 the primary illocution is a declarative: "S₇₈ is saying that S₇₈ promises to go there tomorrow." This is not the illocutionary point of 78, however. S₇₈ is using this primary illocution as a vehicle for a further illocution to be read off the performative verb, namely: S₇₈ reflexively intends the (primary) declarative to be a reason for H₇₈ to believe that S₇₈ undertakes and intends (i.e. S₇₈ promises) to go there tomorrow. There is no further inference to draw, so this is the illocutionary point of 78. By definition, then, the performative clause in 78 communicates an indirect illocution.

What additional evidence is there that performatives are declaratives (in illocution as well as form)? First, there is the obvious similarity between 78 and 79.

79. I promised to go there tomorrow.

Unlike 78, which is in the present tense and has the illocutionary point of a promise, 79 is past tense and has the illocutionary point of a statement (or report) about a promise made in the past. The primary illocution of 79 is "S₇₉ is saying that S₇₉ promised to go there tomorrow." This is not the only parallel with 78, because H₇₉ will interpret 79 (subconsciously, and not in so many words) as 20.

80. S₇₉ reflexively intends the declarative to be a reason for H₇₉ to believe that S₇₉ did undertake and intend to go there tomorrow. There is no further inference to draw, so the illocutionary point of 79 is that S₇₉ did undertake and intend to go there tomorrow.

Note that the undertaking in both 78 and 79 remains to be fulfilled; and although S₇₉ is not actually making the promise in 79 as S₇₈ is in 78, nevertheless, provided all normal cooperative conditions hold, S₇₉ is as much obliged to fulfill the promise reported in 79 as S₇₈ is in 78! The presumption that the primary illocution of explicit performatives is that of a declarative permits a commonsensical account of the similarity and difference between 78 and 79.

Second, there is a distinction between saying Φ and saying that Φ: the former reports locutions, the latter reports statements. Imperatives and interrogatives do not make statements, but declaratives do; compare 81(a-d).

81.

(a_i) Go!
(a_q) What's your name?

(b_i) I said go.
(b_q) I said what's your name?

(c_i) *I said that go.
(c_q) *I said that what's your name?

(d_i) I said that you must go.
(d_q) I said that I want to know your name.

In order to be reported by saying that, the propositional content of imperatives and interrogatives needs to be recast as the declarative in 81(d); this is not the case with a performative because its primary illocution is already that of a declarative; cf. 82, where no (d) row is needed.

82.

(a_d) The beer's cold.
(a_p) I promise to go there tomorrow.

(b_d) I said the beer's cold.
(b_p) I said I promise to go there tomorrow.

(c_d) I said that the beer's cold.
(c_p) I said that I promise to go there tomorrow.

Third, there is a set of adverbials which modify primary illocutionary acts, e.g. honestly, for the last time, seriously, frankly, once and for all, in the first place, in conclusion (Allan 1986, Jackendoff 1972, Schreiber 1972, Sadock 1974). Consider 83:

83. In the first place I admit to being wrong; and secondly I promise it will never happen again.

83 means "The first thing I have to say is that I admit to being wrong; and the second thing I have to say is that I promise it will never happen again." It is clear that 'secondly' denotes a second act of saying, not a second act of promising; from which we may further deduce that 'In the first place' identifies a first act of saying, not a first act of admitting. There is no space to consider more than these three arguments; but the evidence is strongly against the view that explicit performatives are primary illocutions, because primary illocutions are read off the mood which is in most instances readily identifiable from the clause-type.

To top

8. THE INFERENCE SCHEMA FOR CALCULATING ILLOCUTIONARY POINT

Suppose Speaker S₈₄ utters 84 in context C₈₄, where it answers the question What's the time?

84. [Context C₈₄. Question: What's the time? S₈₄ answers:]

It's 7.45.

Many scholars would call S₈₄'s utterance a DIRECT speech act because S₈₄ means exactly and literally "the time now is seven forty-five." Now, suppose that 85 is uttered in context C₈₅.

85. [Context C₈₅. S₈₅ and spouse H₈₅ car-pool to work; they need to leave home by seven forty-five in order to arrive on time.]

It's 7.45.

Although 85 is still a bald-on-record statement of the current time, it is often called an INDIRECT speech act because the speaker of 85, S₈₅, means at least "it is time to leave for work"; and it is quite likely that S₈₅ further implies "hurry up, you're making us late". We reject the binary distinction between direct and indirect as ineffective: partly because there are many degrees of indirectness, but principally because in recognizing that primary illocution is determined by mood every utterance, even those containing explicit performative clauses, has at least a secondary illocution.

According to Searle 1975b:61 an indirect speech act is one in which Speaker performs one illocutionary act but intends Hearer to infer another illocution -- by relying on their mutually shared background information, both linguistic and nonlinguistic. We adopt this notion for the way to process indirectness -- i.e. second and subsequent illocutions in an utterance. Indirectness relies on conversational implicature: there is overwhelming evidence that speakers expect hearers to draw inferences from everything that is uttered -- just as they do from visual and other data perceived and conceived of -- no matter how "direct". It follows that Hearer will begin the inferential process immediately on being presented with the locution. Under the cooperative principle, there is a convention that Speaker has some purpose for choosing to utter U_i in C_i instead of maintaining silence or making some other utterance U_x. Hearer tries to guess this purpose in the course of figuring out the illocutionary point of U_i, and in doing so will consider U_i's relevance to the co-text and setting of the utterance in the light of beliefs about normal behaviour in C_i, beliefs about the speaker of U_i, and the presumed common ground.

Assumption. The INFERENTIAL THEORY of speech acts presents an abstract model of each step necessary in Hearer's reasoning out of Speaker's illocutionary point(s) in making the utterance.

The process of inference is modelled by an inference schema developed out of proposals originally made by Searle 1975b, subsequently refined in the 'speech acts schema' described by Bach and Harnish 1979, and in Allan 1986. There is an assumption that both Speaker and Hearer are normal human beings -- i.e. neither is a genius, a clairvoyant, nor a fool; both know the language and how to use it; and they have the common ground that one can reasonably attribute to such persons in the particular context of utterance. The first step in the process is that Hearer recognizes that Speaker has uttered U and presumably intends to communicate with it; this is recognition of the utterance act. The schema proceeds through the hierarchy of acts (§3) recognizing the locution, the reference, the primary illocution and all subsequent illocutions until the illocutionary point of U has been discovered. To determine the illocutionary point from the primary locution, Hearer invokes the reasonableness condition and seeks some reason for Speaker's primary illocution in the context of utterance. To do so, Hearer checks the semantic properties of the utterance and its constituents in relation to their textual environment, the conversational maxims, and encyclopedic knowledge of many kinds -- including knowledge of the kinds of things that people might be expected to say in such a context and the kinds of reasons that other speakers have had when employing a similar primary illocution. In short, Hearer uses their knowledge of the language and of the use of the language to infer from the primary locution and the circumstances of utterance what the illocutionary point (or points) might be. When Hearer can infer no further illocutions in the utterance the illocutionary point has been reached, and the inference schema for U closes down.⁸

Inference Schema 84 below interprets 84.

Inference Schema 84

The inference schema for 85 shares the same initial five steps with Inference Schema 84, but then it diverges to cope with the different context of utterance.

Inference Schema 85

A serious weakness of speech act theory is the pretence that Speaker's illocutionary intentions can be precisely pinned down. In 85 Speaker tells Hearer the time: are we overestimating Speaker's intention if we assume that s/he thereby implies it is past the time Hearer should have already left for work, hence warning Hearer that s/he is running late and furthermore urging Hearer to hurry? The utterance may merely have been intended to draw Hearer's attention to the time, leaving Hearer to draw whatever conclusion s/he wished. Hearer might be grateful for having the matter brought to his/her attention; get angry with Speaker for interfering; respond by hurrying; respond by suing for divorce because it is the last straw; or responding in innumerable other ways. Speaker discovers only through Hearer's response the perlocutionary effect of the utterance, and thence whether his/her intentions have been realized -- always assuming Speaker has any clear idea what these are!

For a final example, consider 86, which trades on the cooperative requirement that Hearer should have a good reason not to comply with such a request.

86. Why don't you be quiet?

People are not expected to expressly offer reasons for being cooperative, but they are expected to offer reasons for not being cooperative: which is why there is the disjunctive illocutionary point to 86. Notice how the question illocution is carried down through the inference schema to become one of the disjuncts of the illocutionary point of the speech act.

Inference Schema 86

The illocutionary point of any utterance is discovered by an inferential process that attends to Speaker's tone of voice and the context of utterance, knowledge of the language itself and of conversational conventions, and perhaps general encyclopedic knowledge. Speaker knows this and speaks accordingly, aware that Hearer -- as a competent social being and language user -- will recognize the implications of what s/he says. It is not enough to know a language, one must also know how to use it!

To top

9. ON-RECORD AND OFF-RECORD, LITERAL AND NONLITERAL

The study of so-called 'indirect' speech acts has overwhelmingly dealt with requests in various guises. Sadock 1974 identified some exotic species: 'whimperatives' indirectly request (the function of the imperative in Sadock's view) by means of Speaker directly asking a question e.g. Can't you (please) do A? and Do A, will you? (Sadock's analysis is criticized in Allan 1986); using a 'queclarative' Speaker directly questions and indirectly makes an assertion -- Does anyone do A any more? means "Nobody does A any more"; 'requestions' are quiz questions to which Speaker knows the answer, e.g. Columbus discovered America in ...? Blum- Kulka, House, and Kasper 1989:18 identify nine points on an indirectness scale for requests. Their DIRECT STRATEGIES are used when Speaker is dominant:

(i) Imperative: Clean up that mess.
(ii) Performative: I'm asking you to clean up that mess.
(iii) Hedged performative: I would like to ask you to come for a check up.
(iv) Obligation statement: You'll have to move that car.
(v) Want statement: I want you to stop calling me.

What they call CONVENTIONALLY INDIRECT STRATEGIES are Hearer oriented:

(vi) Suggestory formulae: How about cleaning up?.
(vii) Query preparatory (Sadock's whimperatives): Could/Will you clear up the kitchen, please.

Finally, their NON-CONVENTIONALLY INDIRECT STRATEGIES are off-record:

(viii) Strong hint: You've left the kitchen in a dreadful mess.
(ix) Mild hint: I say, it's a bit chilly in here, isn't it? (said when the heating is off and the window open).

The traditional contrast between direct and indirect illocution is muddied by the related contrasts between being ON-RECORD versus OFF-RECORD, and being LITERAL or NONLITERAL. Blum-Kulka et al.'s 'nonconventional indirect' requests could be reclassified "off-record"; the 'direct' and 'conventionally indirect' ones "on- record", because

Definition. An ON-RECORD utterance spells out the message explicitly.

Speaker can be on-record and either more or less indirect; but if Speaker is off-record, s/he is necessarily more indirect. Challenging questions such as 87 superficially seek cessation of or an explanation for the offending act. Indirectly, but nevertheless literally and on-record, Speaker informs Hearer of his/her belief that "the music is very loud."

87. Why play the music so loud?

If Speaker likes loud music, and knows or assumes that Hearer does too, s/he would know why the hifi was playing loud and would not have asked 87 but said something supportive like Great sound! Let's turn up the volume! Therefore, in uttering 87 either (a) Speaker believes that Hearer does not like loud music and is seeking an explanation for this uncharacteristic behaviour or, and more likely, (b) Speaker does not like loud music and is making plain their opinion that "the music is too loud."

For someone who is not very close to Speaker to respond to the invitation in 88 with the bald-on-record refusal No is outright offensive.

88. Do you want come to a movie tonight?

To avoid giving offence we hedge, apologize, prevaricate, and speak off-record, giving reasons for not accepting the invitation or complying with the request. Thus to politely refuse the invitation one says things like 89.

89.

(a) I have to wash my hair.
(b) I'd love to, but my mother's coming to dinner tonight.

Although these might be literally meant on-record statements of Speaker's plans, their illocutionary point is off-record refusal -- a conversational implicature which has to be inferred through additional steps in the inference schema.

A nonliteral utterance such as the sarcastic 90 is an off-record request for Hearer to desist.

90. I'm sure the cat likes you pulling its tail.

What makes it nonliteral is that Speaker does not really mean it as a statement about what the cat likes; in other words, it is the illocution which is nonliteral. However, there is no reason to believe that the literal meaning of 90 is any more direct than an utterance intending the nonliteral meaning of 'bastard' in 91; furthermore, both interpretations are on record.

91. Max is a bastard.

The psycholinguistic evidence (cf. Gernsbacher 1990:89) is that all possible senses of a language expression are activated, and context then suppresses the activation of inappropriate meanings; consequently, there is no reason to believe that the nonliteral meaning of a lexically ambiguous term like bastard takes longer to process than its literal meaning; and therefore no grounds at all for suggesting it is less 'direct'. In contrast to 91, what is nonliteral in 91 is the locution, not the illocution.

Though left to be inferred, all entailments, implicatures, and presuppositions of a proposition within an utterance are communicated; they give rise to indirect illocutions that are often, though not necessarily, deliberately communicated. Suppose Speaker utters 92.

92. Ted's BMW is in the repair shop.

This entails the on-record proposition that "Ted has a BMW" and it is said to be 'indirect' because Speaker does not directly state it. Presumably Speaker literally means "Ted has a BMW" because s/he presupposes its truth (at least, purports to do so). 92 informs Hearer that "Speaker believes Woody has a new job", and Speaker may have made the utterance partly to inform Hearer of this fact.

93. Woody likes his new job.

94 informs Hearer that "Speaker believes that someone phoned", 95 that "Speaker has a married sister" and 96 that "Speaker believes Max has no more than one son and two daughters."

94. Who phoned?
95. My sister's husband is a banker.
96. Max has one son and two daughters.

The public condemnation of the woman at a party who cries out 97 broadcasts on record (a) what Mr Trumpington is (allegedly) doing, (b) a request that he be stopped, (c) Speaker's entailed belief that he will not stop of his own volition.

97. Mrs Trumpington will you please ask your husband to keep his hands off me?!

Off record, Speaker intends not only that Mrs Trumpington condemn her husband for sexual harassment, but that everyone in earshot should do so too, bringing shame to the accused. The phrase keep your hands off is idiomatic and though it is primarily literal, it can also be used with the nonliteral meaning "don't touch".

An on-record utterance spells out the message explicitly; an off-record utterance relies heavily on context and encyclopedic knowledge for its interpretation. Greater indirectness is characteristic of unstated entailments, presuppositions, implicatures, and of nonliteral illocutions. 87 exemplified a literal and on-record challenge Why (must you) do A?; 89 made literal statements with an off-record illocutionary point; 90 a nonliteral statement with the illocutionary point of an off-record request; 91 was an on- record literal assertion containing a nonliteral locution; 92-95 exemplify on-record presuppositions or entailments that are not directly stated; 96 exemplifies an implicature from on-record literally meant constituents; 97 exemplifies both on and off-record literal illocutions, and perhaps a nonliteral on-record one.

To top

10. CULTURAL DIVERSITY

Subject: An application for the position of Post-Doctoral in Linguistics

Sir,
I feel pleasure requesting you for admitting to the Post-Doctoral Course in Linguistics at your esteemed Institution.

(Beginning of a letter from India to the Head of an Australian Linguistics Department, 1996)

All major speech act theorists have ignored cultural diversity, leaving it to empirical studies such as Clyne 1994 or the Cross Cultural Speech Act Realization Project (CCSARP) to investigate (cf. Blum-Kulka, House, and Kasper 1989). The latter project investigated and confirmed cultural differences in strategies for requesting and apologizing in German, Hebrew, Danish, Canadian French, Argentinian Spanish, British, American, and Australian English. For instance, both Americans and Hebrews generally prefer to use on-record Hearer-oriented requests rather than off-record ones, but the second preferences differ: Hebrews prefer Speaker dominant direct requests whereas Americans favour the off-record strategy. Interacting with familiars but not intimates, Slavs, Hebrews, and Germans use direct on-record requests mitigated with e.g. modal particles and diminutives, cf. Wierzbicka 1991; English lacks such devices and therefore uses fewer Speaker dominant strategies. At the present time there is much ongoing research into illocutions which touch on politeness concerns, because it is so very important to avoid inadvertently causing offence in social interaction, especially when Hearer is from another culture.

Different cultural conventions and belief systems in different language communities result in different cooperative conventions. For example, there are different linguistic politeness strategies to reflect the culturally validated perceived roles of Speaker and Hearer, and whatever is being spoken of. Social interactive practices, such as societal attitudes to the roles of men and women, vary greatly across cultures; and, as a result, so do the language conventions people use. Acts of complimenting and thanking are more frequent in Anglo communities than in, say, China or Africa. The Japanese and Chinese reportedly find refusing more offensive than Americans do, and so refuse off-record by using proverbs and impersonals. Preconditions on illocutionary acts defined for one language cannot be expected to be universal. Intercultural miscommunication arises from the assumption that the language strategies appropriate to the delivery of the intended meaning in L_a can be used with equal efficacy in L_b. For instance, direct translation of the English 98 into Polish 99 will, according to Wierzbicka 1991:29, be interpreted as a direct question and not an invitation; the counterpart Polish invitation, she says, is 100. However, other Polish speakers strongly disagree, claiming that 99 is a polite invitation, and 100 more of a suggestion -- both consistent with the English translations.

98. Would you like to go the cinema with me?
99.
100. = "Perhaps we would go to the cinema?"

The English translation of 100 would probably misfire as an intended invitation, because the off-record implication in Polish -- "if I asked you?" -- is lost.

The illocutionary point of any utterance is discovered by an inferential process that attends not only to Speaker's locution but to the conversational conventions that govern the use of language in the particular subculture from which Speaker comes. Misunderstanding can easily arise when Hearer applies the norms from some other subculture, cf. Tannen 1990.

To top

11. MORE THAN ONE ILLOCUTION IN AN UTTERANCE

This section deals with the many sentences which contain more than