Speech act

For the U.S. law, see SPEECH Act.

In the philosophy of language and linguistics, a speech act is an utterance considered as an instance of action in a social context rather than as the mere expression of a proposition. To say "I resign", "I apologise" or "You're fired" is, in suitable circumstances, to perform the very act of resigning, apologising or dismissing, not simply to describe it.^[1] Speech-act theory therefore treats speaking a language as a kind of rule-governed social behaviour in which people make claims, issue orders, ask questions, make promises and so on by means of utterances.

Following J. L. Austin and John R. Searle, many accounts distinguish at least three levels of act in ordinary utterances: the locutionary act of producing a meaningful expression, the illocutionary act performed in saying something (such as asserting, warning, requesting or promising), and the perlocutionary act consisting in its further effects on an audience, such as persuading, amusing or alarming them. Later work has added notions such as metalocutionary acts, which organise or comment on the discourse itself, and has analysed performative utterances and indirect speech acts, in which one kind of act is performed by way of another.

As a systematic theory, the contemporary notion of speech acts originates in Austin's 1955 Harvard lectures published as How to Do Things with Words, and in Searle's subsequent development of explicit rules and taxonomies for illocutionary acts. Historical research has identified important predecessors and parallels, including the later philosophy of Ludwig Wittgenstein, the accounts of "social acts" in the work of Thomas Reid and Adolf Reinach, and early-20th-century linguistic theories of Sprechhandlung ("speech action"), as well as earlier reflections in Hegel on speaking as a form of acting.

Speech-act theory now plays a central role in pragmatics, discourse analysis and communication studies, and has been taken up in a wide range of other fields. In language acquisition it is used to describe how children learn to perform basic communicative acts such as requesting, greeting and protesting; in computer science and information systems it underpins models of human–computer interaction, workflow and multi-agent system communication that treat messages as illocutionary moves creating and discharging commitments; in political science and international relations it informs theories of securitization and the construction of security; and in law, economic sociology and finance it has been applied to the performative role of legal norms and mathematical models in shaping social and economic practices.

History

For much of the early analytic and positivist philosophy of language, language was treated mainly as a vehicle for representing facts and for stating propositions that are either true or false; the action-character of ordinary utterances was largely downplayed or bracketed.^[2] J. L. Austin explicitly criticised this presupposition at the beginning of How to Do Things with Words, noting that philosophers had often assumed that the business of a "statement" is simply to "state some fact".^[1]: 1

A major precursor of speech act theory in the analytic tradition is the later philosophy of Ludwig Wittgenstein, who urged that instead of asking for the abstract meaning of an expression one should look at its use within a practice, and who described speaking a language as engaging in a family of "language-games".^[3] On this view, understanding an utterance requires seeing it as a rule-governed move within a social activity, not merely as a bearer of truth-conditions. Subsequent surveys of speech act theory often take this shift to "meaning-as-use" as part of the background to contemporary work on speech acts and pragmatics.^[4][5]

The contemporary use of the expression speech act is usually traced to the work of the Oxford philosopher J. L. Austin. In his 1955 Harvard lectures, posthumously published as How to Do Things with Words, Austin introduced the distinction between locutionary, illocutionary and perlocutionary acts and drew attention to "performative" utterances such as "I apologize" or "I promise", in which saying something is already doing something.^[6] Austin's analyses encouraged philosophers and linguists to treat a wide range of utterances—requests, orders, apologies, warnings, declarations and so on—as actions performed in accordance with socially recognised rules or "felicity conditions".^[7]

Austin's student John R. Searle further systematised these ideas in Speech Acts: An Essay in the Philosophy of Language (1969) and in later writings, developing explicit rules for the performance of illocutionary acts and proposing influential taxonomies of their types.^[8][9] Subsequent work by Searle and others connected speech act theory with accounts of intentionality, social ontology and discourse structure, and it became a central framework in philosophy of language, pragmatics, communication studies, legal theory and artificial intelligence.^[5][10]

Historical research has shown that Austin and Searle were not the first to analyse what are now called speech acts. In the Scottish common-sense tradition, Thomas Reid distinguished "solitary" mental operations from essentially "social" acts of mind—such as asking, commanding and promising—which require expression to another person and generate obligations between speaker and hearer.^[11][12] Within early phenomenology, Adolf Reinach's 1913 essay Die apriorischen Grundlagen des bürgerlichen Rechtes developed a detailed ontology of "social acts"—promising, claiming, commanding and related performances—that bring about new normative relations when successfully performed.^[13][14] Reinach's work, together with that of Reid and others, has been recognised as an early, systematic speech-act theory developed independently of the later Oxford tradition.^[2]

Comparable ideas were developed in early-20th-century linguistics. The Slovene philologist Stanislav Škrabec analysed performative uses of tense and aspect and discussed utterances in which, as later theorists would say, "saying so makes it so", such as judicial pronouncements and explicit permissions.^[15] The psychologist and linguist Karl Bühler employed the terms Sprechhandlung ("speech action") and Theorie der Sprechakte ("theory of speech acts") in his 1930s work on the functions of language, especially in Sprachtheorie (1934), which has been read as an important forerunner of later speech act theorising.^[16][17][18]

Beyond these relatively direct predecessors, some scholars have also emphasised the roots of speech-act thinking in classical German philosophy. In his lectures on the philosophy of right, Georg Wilhelm Friedrich Hegel describes speaking as a form of acting (sprechen ist so nicht bloß sprechen, sondern handeln), linking utterance to the actualisation or transformation of social reality.^[19] These earlier reflections on "social" or "performative" uses of language are brought together in historical reconstructions of speech act theory that place Austin and Searle within a broader, multi-stranded tradition.^[2]

Types of speech act

Main articles: locutionary act, illocutionary act, perlocutionary act, and metalocutionary act

Following J. L. Austin's distinction in How to Do Things with Words, many philosophers and linguists analyse ordinary utterances as involving at least three kinds of act: a locutionary act (producing a meaningful linguistic expression), an illocutionary act (performing an action such as asserting, questioning, or promising in saying that expression), and often a perlocutionary act (bringing about further effects on an addressee, such as convincing or alarming them).^[20][21] Later theories refine this framework but usually retain some version of the threefold distinction.^[22]

In this tradition, speech acts can be described on multiple, partly overlapping levels:

1. A locutionary act is the act of producing a particular linguistic expression with a certain phonological form, syntactic structure and conventional meaning—the act of saying something meaningful in a language.^[20][23]
2. An illocutionary act is the act a speaker performs in saying something, characterised by its illocutionary force: for example, asserting that something is the case, asking a question, issuing an order, giving a warning, or making a promise.^[20][24]
3. A perlocutionary act is the act a speaker performs by saying something, namely the further effects an utterance has on an audience, such as persuading, discouraging, amusing, frightening, or getting someone to perform a certain action, whether those effects are intended or not.^[20][23] For example, if a speaker says "I'm hungry" and the hearer responds by preparing food, the hearer's reaction can be described as part of the perlocutionary effect of the original utterance.
4. A metalocutionary act is a speech act that comments on or organises the discourse itself (for instance, marking topic boundaries, highlighting quoted material, or signalling the start or end of a turn) rather than further advancing its subject matter. In prosody and orthography, metalocutionary functions have been analysed as the use of intonation and punctuation to highlight or delimit portions of an utterance.^[25][26]

Illocutionary acts

The notion of an illocutionary act is central in speech-act theory, since it corresponds to the basic kind of social action a speaker performs by means of an utterance. Austin's examples include acts of stating, warning, ordering, and promising, each governed by appropriate "felicity conditions" that must obtain if the act is to succeed (for instance, that the speaker has the requisite authority to pronounce a couple married).^[20][21]

Searle develops Austin's insights by treating "speech act" and "illocutionary act" largely interchangeably and by offering a systematic taxonomy of basic illocutionary types (such as assertives, directives, commissives, expressives and declarations) characterised in part by the psychological states they express (beliefs, desires, intentions, and so on).^[24][27] Whereas Austin tends to emphasise the conventional procedures associated with illocutionary acts, Searle and later authors highlight the role of underlying mental states and inferential recognition in individuating and understanding them.^[22][21]

Perlocutionary acts

While illocutionary acts are primarily defined in terms of the speaker's communicative intention and the conventional force of an utterance, perlocutionary acts concern the actual effects that utterances have on hearers and other addressees. These effects can involve changes in belief, emotion or behaviour: an assertion may convince someone, a warning may alarm them, a request may prompt them to act, and so on.^[20][23] Because perlocutionary effects depend on the audience and context, the same utterance may have different, even unintended, consequences for different hearers.^[21]

Performative speech acts

A particularly clear class of illocutionary acts is found in performative utterances, where the act named by a verb is performed in uttering a sentence containing that verb in an appropriate form and context. Typical examples include "I promise to pay you back", "I apologise for being late", "I sentence you to ten years' imprisonment", or "I hereby name this ship Sojourner". In such cases the utterance does not merely describe a promise, apology, sentence, or naming; it constitutes one, provided the relevant felicity conditions are satisfied.^[20][21]

Austin argued that performative utterances are not aptly evaluated as true or false, but rather as "happy" (felicitous) or "unhappy" (infelicitous), depending on whether those conditions are met—for instance, whether the speaker is authorised to perform the act, whether the procedure is correctly executed, and whether the participants have the appropriate intentions.^[20] Subsequent work has refined this idea by distinguishing explicit performatives (such as "I promise…") from cases where the same act is performed without using a corresponding performative verb (as when saying "I'll be there at six" functions as a promise in the right context).^[22][21]

Linguists sometimes use the term performative speech act more narrowly for utterances that contain a first-person singular present-tense performative verb and pass the so-called "hereby test" ("I hereby promise/apologise/declare…"). Expressions like "I intend to go" or "I hope to go", although they use first-person present-tense verbs, normally report mental states rather than perform conventional acts of promising or vowing, and so are not standardly treated as explicit performatives.^[23][28]

Indirect speech acts

In many cases, speakers perform speech acts indirectly: they use one sentence type to perform an illocutionary act that is different from the one conventionally associated with that form. For example, the question "Can you close the window?" is typically used not to inquire about the hearer's ability but to make a request, and the declarative "It's cold in here" may be understood, in context, as prompting someone to close a window or turn up the heat.^[29][28]

Indirect speech acts are especially common for making requests and refusals. When asked "Would you like to meet for coffee?", a reply such as "I have class" is ordinarily taken as a polite rejection rather than as a bare report about one's schedule.^[23] Many languages make routine use of conventionalised indirect request forms (for example, English "Could you…?", "Would you mind…?"), which have become standard ways of mitigating the social imposition associated with directives.^[30][31] They may also rely on non-verbal cues: for instance, fieldwork on Jordanian shops and streets describes vendors using music, scents and other environmental signals, alongside verbal formulas, to attract customers and invite purchases without issuing explicit commands.^[32]

Such cases pose a classic problem for theories of communication: how do hearers reliably identify the intended illocution when the literal sentence meaning points to another act? In a seminal proposal, Searle argued that indirect speech acts can be derived by a sequence of inferences based on Grice's Cooperative Principle and conversational maxims: the hearer reasons from the literal meaning and contextual information to the hypothesis that a request, offer, refusal, and so on, is being performed.^[29][33] Subsequent work in pragmatics, politeness theory and experimental psycholinguistics has questioned whether a single inferential pattern can account for all indirect speech acts and has proposed alternative models for how hearers recover indirect meanings.^[34][35][36]

More generally, speakers need not explicitly use verbs such as "apologise", "promise" or "praise" in order to perform those actions; whether an utterance counts as an apology, a promise or a compliment depends on its conventional form and on how it is understood in context.^[23][28]

Examples

Speech acts are commonplace in everyday interactions and are important for communication, as well as present in many different contexts. Examples of these include:

• "You're fired!" expresses both the employment status of the individual in question, as well as the action by which said person's employment is ended.^[37]
• "I hereby appoint you as chairman" expresses both the status of the individual as chairman, and the action that promotes the individual to this position.^[38]
• "We ask that you extinguish your cigarettes at this time, and bring your tray tables and seatbacks to an upright position." This statement describes the requirements of the current location, such as an aeroplane, while also issuing the command to stop smoking and to sit up straight.
• "Would it be too much trouble for me to ask you to hand me that wrench?" functions to simultaneously ask two questions. The first is to ask the listener if they are capable of passing the wrench, while the second is an actual request.
• "Well, would you listen to that?" acts as a question, requesting that a listener heed what is being said by the speaker, but also as an exclamation of disbelief or shock.^[39]

Felicity conditions and misfires

In How to Do Things with Words, Austin introduced the notion of felicity conditions for the conventional procedures he took to underlie many performative utterances and other illocutionary acts.^[20][21] On his account, the success or "happiness" of a speech act depends not only on the conventional meaning of the words uttered but also on background features of the context of utterance: there must be an accepted procedure for performing the act, the circumstances and participants must be appropriate to that procedure (for instance, a legally authorised official pronouncing a couple married), and the procedure must be carried out correctly and completely, with participants having the requisite thoughts, feelings and intentions and, where relevant, engaging in the expected subsequent conduct.^[20][40] Felicity conditions are thus pragmatic constraints on when an utterance of a given type counts as a valid performance of a particular kind of act, rather than as a mere emission of words or as a statement that can simply be evaluated for truth or falsity.^[21][41]

Austin calls failures of felicity infelicities and divides them into two main classes, misfires and abuses.^[20][21] In cases of misfire the purported speech act does not come off at all: the conventional procedure is not properly invoked or executed, so that, although words are spoken, no marriage is contracted, no bet is laid, no christening or declaration succeeds.^[20][40] This may happen, for example, because the speaker lacks the relevant authority, because some presupposed condition is not met (such as there being an appropriate object or addressee), or because participants fail to complete the required sequence of actions. By contrast, abuses are cases in which the conventional act is successfully performed but is defective because some sincerity or subsequent-conduct condition is violated—for instance, when a speaker promises without intending to keep the promise, makes a bet without intending to pay, or issues an apology without regret.^[20][21] In such cases there is still a promise, bet or apology, but it is an insincere or otherwise defective one; sincerity is treated as a paradigm felicity condition that an act of a given type ought, but need not, satisfy.^[21][40]

Searle generalised Austin's ideas by analysing felicity conditions as sets of constitutive rules associated with each basic type of illocutionary act.^[24][21] For a paradigmatic case such as promising, he distinguishes four kinds of condition: a propositional content condition restricting what may be promised; preparatory conditions concerning the relationship between speaker, hearer and situation (for example, that the hearer prefers the promised action to its absence, and that the act is not already certain to occur); a sincerity condition requiring the speaker to intend to carry out the promised action; and an essential condition according to which the utterance counts as undertaking an obligation.^[24][41] Violations of content or preparatory conditions are typically treated as yielding failed or infelicitous attempts to perform the act (a kind of misfire), whereas violations of the sincerity condition generate insincere but nonetheless binding promises, orders or apologies (a kind of abuse).^[21][41]

Later work extends the idea of felicity conditions beyond explicit performatives to speech acts more generally, including assertions, questions and indirect speech acts, and links them to norms governing what speakers may do with words in different social settings.^[41][40] On this view, felicity conditions articulate the normative background that allows utterances to create or modify obligations, entitlements and other elements of social reality, and they can be used to explain how power relations, institutional roles and patterns of uptake affect which acts can be successfully performed.^[21][42] Contemporary discussions debate how far such conditions can be captured in finite lists for each illocutionary type and to what extent they should instead be seen as defeasible, context-sensitive constraints on when speakers are entitled to perform particular acts and when hearers are entitled to treat those acts as successful and binding.^[21][42]

In language development

In 1975 John Dore proposed that children's utterances were realizations of one of nine primitive speech acts:^[43]

1. labelling
2. repeating
3. answering
4. requesting (action)
5. requesting (answer)
6. calling
7. greeting
8. protesting
9. practicing

Formalization

There is no agreed formalization of Speech Act theory. In 1985, John Searle and D. Vandervecken attempted to give some grounds of an illocutionary logic.^[44] Other attempts have been proposed by Per Martin-Löf for a treatment of the concept of assertion inside intuitionistic type theory, and by Carlo Dalla Pozza, with a proposal of a formal pragmatics connecting propositional content (given with classical semantics) and illocutionary force (given by intuitionistic semantics). Up to now the main basic formal applications of speech act theory are to be found in the field of human–computer interaction in chatboxes and other tools. Recent work in artificial intelligence proposes a Bayesian approach to formalize speech acts ^[45]

In computer science

In computer science and information systems, speech act theory has been used to model human–computer interaction, office work, multi-agent system communication and other forms of computer-mediated interaction by treating messages as illocutionary actions that change the social state of an interaction rather than merely transmitting data.

Human–computer interaction and dialogue systems

Early work on computational models of speech acts in human–computer interaction proposed representing dialogue as a sequence of illocutionary moves, with explicit changes of conversational state associated with each move. In 1991, Morelli, Bronzino and Goethe described a computational speech-act model of human–computer conversations for medical decision support, in which user inputs and system replies are typed as requests, assertions, confirmations and other speech-act categories that drive the underlying expert system.^[46]

Later work has used speech-act based models to support automated classification and retrieval of conversations: for example, Twitchell et al. modelled conversations in e-mail and chat as sequences of speech acts in order to classify threads and retrieve segments relevant to particular tasks.^[47]

More recent natural-language processing research treats speech-act recognition as a text classification problem. Qadir and Riloff, for example, train statistical classifiers to recognise sentences in online message boards as instances of Searle's main speech-act categories (commissives, directives, expressives and representatives).^[48] Related work has applied machine-learning methods to classify speech acts in online chat and other forms of computer-mediated communication.^[49]

Conversation for action and workflow modelling

Another influential application of speech act theory is the conversation for action framework developed by Terry Winograd and Fernando Flores in Understanding Computers and Cognition: A New Foundation for Design.^[50] They analyse everyday coordination of work as networks of conversations in which participants make requests, offers, promises and declarations, and represent these conversations using a state-transition diagram that tracks the illocutionary status of each commitment (for example, whether a request has been accepted, fulfilled or declined).^[51] In this view, a computer process can track the social state of a transaction—such as which commitments have been made or discharged—even when it does not model in detail the external world that the commitments concern.

The conversation-for-action model has influenced computer-supported cooperative work, workflow management and business process modelling. For example, Medina-Mora et al. propose action workflow as an office-automation architecture in which work is coordinated through structured conversations for action,^[51] and Auramäki, Lehtinen and Lyytinen use a speech-act based office modelling approach (SAMPO) to analyse office activities as chains of commitments created and modified by speech acts.^[52] This line of work is sometimes described as the Language/action perspective on information systems design.^[53]

Rules and protocol design

In specifying communication protocols for distributed and multi-agent systems, computer scientists have drawn on John Searle's distinction between regulative and constitutive rules.^[54] Regulative rules prescribe or constrain behaviour in an activity that could in principle exist without them (for example, traffic regulations), whereas constitutive rules do not merely regulate but help to define an activity, such as the rules of games or institutional practices.^[55][56] Within the language–action and multi-agent systems traditions, interaction protocols are often described as constitutive rules that create and shape social realities such as commitments, permissions and institutional facts, rather than as mere constraints on message passing.^[57]

Multi-agent systems

In multi-agent systems, communication between software agents is commonly modelled using speech-act labels that express the intended illocutionary force of a message. A message with the performative inform, for example, is understood as an attempt to add some content to the recipient's knowledge base, whereas a performative such as request or query asks another agent to perform an action or provide information. Early agent communication languages such as KQML and the FIPA Agent Communication Language define sets of performatives (such as "inform", "request" and "query") together with a formal semantics inspired by Searle's analysis of speech acts.^[58][59]

The semantics of KQML and FIPA ACL are often described as mentalist or psychological, because they interpret communicative acts in terms of the beliefs, desires and intentions that agents are presumed to have.^[60] Munindar P. Singh has argued that such mentalist semantics are ill-suited to open systems, and has instead advocated a social semantics in which communication creates and manipulates publicly observable social commitments between agents, without making strong assumptions about internal mental states.^[61][62] Andrew J. I. Jones and co-authors have likewise criticised psychological approaches to agent communication, arguing for semantics grounded in social and institutional facts.^[63] A later collection of manifestos on agent communication reports a broad consensus in favour of commitment-based social semantics for open multi-agent systems and questions the adequacy of the original FIPA ACL semantics for such settings.^[64]

Other uses in technology

• An office can be seen as a system of speech acts. The abbreviation SAMPO stands for Speech-Act-based office Modeling approach, which "studies office activities as a series of speech acts creating, maintaining, modifying, reporting, and terminating commitments".^[65]
• Speech act profiling and related techniques have been used to detect deception in synchronous computer-mediated communication, for example by analysing the distribution of different speech-act types in chat and instant-messaging conversations.^[66]
• Automatic speech-act classification has been applied in a variety of natural-language processing tasks, including the detection of questions, answers and other conversational roles in online discussion fora and chat logs.^[67][68]

In political science

In political science, the Copenhagen School adopts speech act as a form of felicitous speech act (or simply 'facilitating conditions'), whereby the speaker, often politicians or players, act in accordance to the truth but in preparation for the audience to take action in the directions of the player that are driven or incited by the act. This forms an observable framework under a specified subject matter from the player, and the audience who are 'under-theorised [would] remain outside of the framework itself, and would benefit from being both brought in and drawn out.'^[69] It is because the audience would not be informed of the intentions of the player, except to focus on the display of the speech act itself. Therefore, in the perspective of the player, the truth of the subject matter is irrelevant except the result produced via the audience.^[70]

The study of speech acts is prevalent in legal theory since laws themselves can be interpreted as speech acts. Laws issue out a command to their constituents, which can be realized as an action. When forming a legal contract, speech acts can be made when people are making or accepting an offer.^[71] Considering the theory of freedom of speech, some speech acts may not be legally protected. For example, a death threat is a type of speech act and is considered to exist outside of the protection of freedom of speech as it is treated as a criminal act.

In economic sociology

In a sociological perspective, Nicolas Brisset adopts the concept of speech act in order to understand how economic models participate in the making and the spreading of representations inside and outside of the scientific field. Brisset argues that models perform actions in different fields (scientific, academic, practical, and political). This multiplicity of fields induces a variety of felicity conditions and types of performed actions. This perspective is a criticism of the essentialism of philosophical modelling studies.^[72] This approach is largely inspired by the work of Pierre Bourdieu^[73] and Quentin Skinner.

In finance, it is possible to understand mathematical models as speech acts: in 2016 the notion of "financial Logos" was defined as the speech act of mathematical modelling of financial risks. The action of the financial Logos on financial practices is the framing of financial decision-making by risk modelling.^[74]

See also

• Analogy – Cognitive process of transferring information or meaning from a particular subject to another
• Cooperative principle – Pragmatics of conversational communication
• Dialog act – Type of speech act
• Direction of fit
• Discourse-completion task
• Entailment (pragmatics) – Concept in linguisticsPages displaying short descriptions of redirect targets
• Implicature – Information conveyed verbally yet not literally
• How to Do Things with Words (J.L. Austin's book) – English philosopher (1911–1961)
• Metaphor – Figure of speech of implicit comparison
• Performative verb – Verbs carried out through being uttered
• Phatic expression – Utterances which primarily serve a social function
• Presupposition – Assumed context surrounding an utterance
• Politeness theory – Social and linguistic theory of politeness
• Relevance theory – Theory of cognitive linguistics