Speech Acts

Austin distinguished three acts performed in any utterance. The locutionary act is the act of saying — uttering certain sounds with certain meanings. The illocutionary act is what one does in saying — promising, warning, asserting, requesting. The perlocutionary act is what one does by saying — convincing, frightening, persuading, amusing. Saying "the door is open" (locution) may count as a request to close it (illocution) and may cause the hearer to close it (perlocution). The illocutionary level is what speech-act theory primarily targets — the conventional force of utterance.

Austin observed that performatives cannot be true or false but can succeed or fail. He proposed felicity conditions for success. Conventional procedure: there must be an accepted procedure with conventional effect (e.g., a wedding requires officiant, vows). Authority: participants must be appropriate (a clergy member can marry; a stranger cannot). Sincere intent: speakers must actually intend the act (a sham promise is infelicitous). Subsequent conduct: speakers must follow through. Searle (1969) refined these into preparatory, sincerity, propositional content, and essential conditions. Each speech act has its own felicity profile.

John Searle, in "A Classification of Illocutionary Acts" (Language in Society, 1976), distinguished: Representatives (assertions, claims, reports — committing speaker to truth of proposition); Directives (requests, commands, questions — getting hearer to do something); Commissives (promises, threats, offers — committing speaker to future action); Expressives (apologies, thanks, congratulations — expressing psychological state); Declarations (christening, firing, declaring war — bringing about institutional change by saying so). Each class has distinctive direction-of-fit (word-to-world or world-to-word) and sincerity conditions. The taxonomy guides linguistic analysis and dialogue-system design.

An indirect speech act performs one illocutionary act by means of another. "Can you pass the salt?" is literally a question about ability; it functions as a request. Searle's "Indirect Speech Acts" (1975) explained the phenomenon through felicity conditions — questioning a preparatory condition (the hearer's ability) implies the request. Conventionalized indirect requests ("Could you...", "Would you mind...") are common in English and serve politeness. Languages vary in conventional indirectness — Japanese is highly indirect; Hebrew often more direct. Translation between these conventions is a frequent pragmatic pitfall.

Penelope Brown and Stephen Levinson's Politeness: Some Universals in Language Usage (1987) treated many speech acts as inherently face-threatening — requests threaten the hearer's negative face (autonomy), criticism threatens positive face (approval). Speakers mitigate threats through politeness strategies: bald-on-record (no mitigation), positive politeness (emphasizing solidarity), negative politeness (emphasizing autonomy, deference, hedging), off-record (indirect hints), or no FTA at all. The framework predicts cross-cultural patterns and gender differences. It has been critiqued and extended (Eelen, 2001; Watts, 2003) but remains foundational.

Dialogue systems use dialogue act tagging — labeling each utterance with its speech-act type. The DAMSL scheme (Allen and Core, 1997) and SWBD-DAMSL (Switchboard, 2000) provide hierarchies for spontaneous conversation. Modern systems (chatbots, virtual assistants) classify user utterances by act (request, question, confirmation) and select appropriate response acts. Reinforcement-learning dialogue agents (DeepMind 2018, Meena 2020, ChatGPT 2022) implicitly handle speech acts but lack explicit illocutionary representations. Hybrid systems (Rasa, Microsoft Bot Framework) combine intent classification (a speech-act surrogate) with response generation.

Austin's framework challenged the descriptivist view dominant in early analytic philosophy — that meaning is fundamentally about truth conditions. Speech-act theory shows much language use cannot be assessed for truth — promises, apologies, queries do not describe states of affairs. The framework links to Wittgenstein's later philosophy (language games, Philosophical Investigations 1953), Grice's pragmatics (1975), and Robert Brandom's inferentialism (Making It Explicit, 1994). Judith Butler's Excitable Speech (1997) extended speech-act theory to politics — hate speech, performative identity, sovereignty. The philosophical reach is wide.