Speech Act Theory: Austin & Searle's Pragmatics Explained

Indirect speech acts occur when speakers veil their true intentions behind polite or neutral formulations. Examples include 'I was wondering if maybe we could take care of it here' (veiled bribe), 'If you could pass the guacamole, that would be awesome' (polite request), 'We're counting on you to show leadership' (solicitation for donation), 'Would you like to come up and see my etchings?' (sexual come-on), and 'Nice store you got there' (veiled threat). Both parties understand the true meaning despite the indirect formulation.

A speech act is the action performed by a speaker with an utterance, such as promising when saying 'I will be there at 6:00'. Speech acts include requesting, commanding, questioning, and informing. Indirect speech acts occur when syntactic structure differs from function—for example, 'Can you pass the salad?' is interrogative in structure but functions as a request. This indirectness is often used for politeness, as direct requests can threaten the listener's face by assuming social power.

Speech acts can be direct or indirect. In direct speech acts, the intention is clear in the message—for example, when a speaker says 'I congratulate you' with the intention of congratulating someone. In indirect speech acts, the intention is not explicit in the message—for example, when a speaker says 'I congratulate you' with a mocking tone, the metamessage (the implicit message transmitted through cues like tone, gestures, and word choice) conveys sarcasm rather than sincere congratulations.

Politeness in communication involves managing social interactions to preserve both one's own and others' 'face' or social image, achieved through face-threatening acts (direct speech that risks damaging someone's image) and face-saving acts (indirect speech that protects social standing), with positive face representing the desire to be liked and negative face representing the desire for autonomy and not being imposed upon.

A speech act is direct when there is a direct relationship between structure and communicative function. For example, 'Did you eat pizza?' uses interrogative form to function as a question, and 'Close the door' uses imperative form as a command. Direct speech acts explicitly express intended meaning. An indirect speech act occurs when form does not directly correspond with function, such as using interrogative form to make a request ('Can you pass me the salt?') rather than ask a question. The main reason for using indirect speech acts is that they are generally considered more polite and formal in many cultures. For example, 'Could you please open the door for me?' sounds more polite than 'Open the door.'

Grice's Cooperative Principle forms the theoretical foundation of the Conversational School. It states that successful communication requires speakers to be cooperative and committed to providing maximum information. When speakers violate this principle, listeners infer meaning through 'conversational implicature' - meaning beyond literal words. For example, 'Xaled ate some bread' implies Xaled did not eat all the bread because the speaker violated the maxim of quantity. Grice proposed four maxims: Quantity (provide enough information), Quality (be truthful), Relation (be relevant), and Manner (be clear). These principles explain how meaning is constructed through contextual inference rather than linguistic structure alone.

Paul Grice proposed that conversation is an act of cooperation between allies, governed by four maxims: Quality (truthfulness), Quantity (appropriate information), Relevance (contextual appropriateness), and Manner (clarity). This principle operates unconsciously. Grice identified two types of conversational implicature: (1) Resolving ambiguity—when statements are literally ambiguous, we infer meaning based on cooperative assumptions (e.g., 'I saw Juan with a woman' implies an affair); (2) Causal inference—we assume the first mentioned event caused the second (e.g., 'She fell and broke her hip' implies falling first).

The Cooperative Principle, formulated by philosopher Paul Grice, states that successful communication assumes speakers are trying to communicate effectively by following four maxims: Quality (tell the truth), Quantity (provide necessary information without excess), Relevance (stay on topic), and Manner (avoid obscurity). When speakers appear to violate these maxims, listeners infer additional unstated meanings called conversational implicatures, which allow us to understand the true intended meaning of statements that seem incomplete or indirect, such as answering 'yes' to 'do you know what time it is?' when the real answer is 'no' (you don't know the time).

Paul Grice proposed that successful communication requires speakers and listeners to cooperate through four maxims: Quantity (provide sufficient but not excessive information), Quality (provide truthful and verified information), Relation (stay on topic and maintain relevance), and Manner (be clear, orderly, and avoid ambiguity). When speakers violate these maxims, listeners can detect violations through conversational implicature (sezdirim), which refers to implied meanings beyond literal words. For example, when someone mentions a conflict of interest, they may be implying they cannot fulfill a request without explicitly stating it. This pragmatic understanding allows listeners to interpret communicative intentions even when speakers do not follow the maxims explicitly.

Paul Grice's cooperative principle (1975) establishes that successful communication requires speakers to cooperate by contributing appropriately to the conversation's purpose. This creates mutual trust between communicators. The four maxims guide cooperation: quantity (be informative but not excessive), quality (tell truth with evidence), relation (stay relevant), and manner (be clear and orderly). When maxims appear violated, listeners infer hidden meanings through conversational implicature. Implicature divides into conventional (automatic from language, like 'but' implying contrast) and conversational (context-dependent, cancellable). Scalar implicature involves quantity maxims (e.g., 'some' implies not all). Flouting deliberately violates maxims for strategic effects, requiring listeners to infer alternative meanings.

This section explains how researchers connected language and reasoning by encoding mental states—goals, intentions, and plans. A plan is a sequence of actions transforming an initial state to a goal state, with actions having preconditions and effects. Speech acts (informing, requesting, promising) were formalized as actions performed by speaking. This enabled plan-based conversational agents that listen, infer intentions, identify obstacles, and plan responses. This model explains phenomena like helpful responses in fragmented conversations (e.g., 'black beans' → 'aisle three') and indirect speech acts (e.g., 'can you pass the salt' as a request).

Speech act theory describes rituals performed through words, such as marriage, baptism, and divorce, where the action is completed by the spoken word. In conversational design, this theory helps understand how people structure requests and orders. The theory reveals that communication goes beyond syntactic structure - for example, in Portuguese, requests may not use explicit markers like 'please' but are understood through tone of voice and linguistic resources. This knowledge helps designers create more natural and effective conversational interactions by understanding the pragmatic elements of human communication.

This section presents Vinograd and Flores's application of speech act theory to computing. Drawing from Austin and Searle, they discovered that language has not only denotative and descriptive character but also a social dimension. Linguistic acts are characterized not by truth or falsity but by the relational commitments they create between speakers. A promise creates commitment, and human life is characterized by a network of such commitments that create an 'expectability' of the future, though not certainty, because commitments can be broken. When commitments are broken (e.g., someone doesn't show up to an appointment), they are repaired through linguistic practice—calling to ask why and rescheduling. Vinograd and Flores built systems to manage these commitment networks, going beyond communication to capture the meaning of communication in human relationships. This represents a fundamental shift from viewing communication as mere information transfer to understanding it as the fabric of social existence.

Evaluation reveals significant challenges: third-party experts show high precision but low recall, while real-time participants show even lower performance (55% precision, 33% recall). This discrepancy suggests real-time awareness affects annotation behavior. Beyond meeting highlights, speech act analysis improves NLU in conversational agents, enables efficient dialogue management, automates to-do list creation from chat messages, analyzes call center interactions, and supports politeness detection for style transfer. These diverse applications demonstrate the framework's versatility across different conversational contexts.

Effective conversational agents must understand human interaction patterns through multiple theoretical lenses. Speech Act Theory distinguishes locutionary acts (literal meaning), illocutionary acts (implied social functions), and perlocutionary acts (effects on listeners). Dialogue act classification identifies abstract conversation moves like statements, acknowledgments, opinions, and questions, enabling agents to recognize context-appropriate responses. Research mines customer care call logs to derive dialogue strategies: offering choices for helpful users, allowing venting time for angry customers, and providing clarification for confused users. These patterns enable agents to adapt communication style based on detected personality traits, providing appropriate levels of detail and abstraction. The inherent uncertainty of emotion and intent detection (~66-70% accuracy) requires strategies like confirmation checkpoints and conversation repair mechanisms, drawing parallels to human communication where we make assumptions about others' knowledge and adjust based on feedback.

A performative utterance is a statement that accomplishes what it states—statements that are acts rather than merely descriptions. Examples include 'I now pronounce you husband and wife' (which makes a marriage official) and 'I forgive you' (which actually forgives). Unlike classical language analysis that treats discourse as autonomous objects, performative utterances prove that statements cannot be separated from their speaker and context.

Performatives are statements that accomplish what they declare simply by being uttered. In legal contexts, performatives include marriage ceremonies (where declarations transform relationships), graduation ceremonies (where declarations change status), and meeting closures. Humans inhabit multiple worlds: the physical world (trees, mountains), the ideal world (numbers, theorems), and the social world (money, property, marriages, laws). The social world is created by humans through culture and has performative power - it modifies reality without materially changing it.

Legal analysis distinguishes between constative discourse (describing the world, communicating opinions) and performative discourse (changing the world, influencing environment). Incitement to hate, violence, or discrimination is a performative speech act where the speaker accomplishes an action through speech. Belgian judges examine speaker intention and context, not just content. The VLAM case condemned deliberate incitement to violence against immigrants, not extremist discourse itself. The 1995 negationism law targets the performative element of lying with bad faith to exonerate Nazism, protecting historical research from abuse.

J.L. Austin distinguished between performative utterances (which perform actions) and constative utterances (which state facts). Constatives can be true or false (e.g., 'It is raining'), while performatives cannot be evaluated for truth value—they are either felicitous (appropriate) or infelicitous (inappropriate). Performative verbs often take special modifiers like 'hereby' ('I hereby order you to leave'), and examples include promising, ordering, marrying, declaring war. Austin noted that performative verbs can sometimes take indicative mood without 'hereby' (e.g., 'Leave the room!').

Verdicts of guilt are performative speech acts that impose liability, not merely declarations of fact. When jurors say 'the defendant is guilty,' they perform an action that changes the defendant's legal status, requiring institutional authority and procedural conditions that ordinary speakers lack. This insight from Judith Thompson's work highlights the crucial connection between speech acts and legal verdicts, suggesting that declaring guilt is fundamentally different from simply asserting that someone is guilty.