Speech act
Based on Wikipedia: Speech act
Words That Do Things
"You're fired."
Two words. In the right circumstances—spoken by someone with authority, directed at an employee, in a professional setting—these words don't merely describe a termination. They are the termination. The moment they leave the speaker's mouth, reality shifts. A person who was employed is now unemployed. The words didn't report an event; they created one.
This is the central insight of speech act theory, and once you see it, you can't unsee it. We tend to think of language as a tool for describing the world—for making statements that are either true or false, like "the cat is on the mat" or "water boils at 100 degrees Celsius." But an enormous portion of what we actually do with language isn't description at all. It's action.
When a judge pronounces a sentence, when a bride says "I do," when a monarch declares war, when you apologize to a friend—these utterances don't report pre-existing facts. They bring new facts into existence. Speaking, in these cases, is a form of doing.
The Philosopher Who Noticed
The person who turned this observation into a systematic theory was John Langshaw Austin, an Oxford philosopher who delivered a series of lectures at Harvard in 1955. These lectures, published posthumously as How to Do Things with Words, changed how philosophers and linguists think about language.
Austin began by poking fun at his fellow philosophers. They had spent centuries treating language as if its primary job were to make true or false statements about the world. But this obsession with truth and falsity, Austin argued, had blinded them to something obvious: much of what we say can't be evaluated as true or false at all.
Consider "I apologize." Is this statement true? Is it false? The question doesn't quite make sense. What matters isn't whether it accurately describes some independent state of affairs, but whether it succeeds as an apology. Did the speaker mean it? Was there something to apologize for? Did the recipient accept it? These are the relevant questions, and they have nothing to do with truth conditions.
Austin called these utterances "performatives" because saying them performs an action. And he contrasted them with "constatives"—ordinary descriptive statements that can be true or false. But as his lectures progressed, Austin realized the distinction was harder to maintain than he'd thought. Even seemingly pure descriptions ("The cat is on the mat") are doing something—they're asserting, informing, claiming. The line between describing and performing began to blur.
Three Acts in Every Utterance
Austin's solution was to analyze utterances as having multiple layers, multiple things happening simultaneously. He identified three kinds of acts that occur whenever we speak:
First, there's the locutionary act. This is simply the act of producing a meaningful expression—combining sounds into words, words into sentences, according to the rules of a language. When you say "It's cold in here," you're performing a locutionary act: making certain sounds that, in English, carry certain meanings.
Second, there's the illocutionary act. This is what you're doing in saying something—the social action your utterance performs. "It's cold in here" might be an assertion (stating a fact), a complaint, a request (could someone close the window?), or even a warning. The illocutionary act is the force behind the words, the punch they carry.
Third, there's the perlocutionary act. This is what you accomplish by saying something—the effects your utterance has on your audience. If saying "It's cold in here" causes someone to close the window, that's the perlocutionary effect. If it makes them feel guilty for not noticing sooner, that's another perlocutionary effect. These consequences might be intended or unintended, predicted or surprising.
The same string of words can have vastly different illocutionary forces depending on context. "I'll be there at six" can be a prediction, a promise, or a threat. The literal meaning—the locutionary content—stays the same. But what the speaker is doing with those words varies enormously.
When Words Succeed and When They Fail
Since performative utterances are actions rather than descriptions, Austin argued they should be evaluated differently. Instead of asking whether they're true or false, we should ask whether they're "felicitous" or "infelicitous"—whether they succeed or fail as actions.
What makes a performative succeed? Austin identified several conditions, which he called "felicity conditions." The right procedure must exist and be invoked. The circumstances and persons must be appropriate. The procedure must be executed correctly and completely. And the participants must have the right intentions and follow through on any obligations created.
Consider the act of marrying someone. In most Western legal systems, certain people (clergy, judges, licensed officiants) have the authority to marry couples. A ceremony must be conducted. Vows must be exchanged. Documents must be signed. If any of these conditions fails—if an unauthorized person performs the ceremony, if the couple is already married to others, if they're performing in a play rather than actually getting married—the performative misfires. The words "I now pronounce you married" have been spoken, but no marriage has occurred.
This framework reveals something important: the power of words depends entirely on social institutions. "I hereby declare war" is just noise when spoken by an ordinary citizen. But when spoken by someone with constitutional authority to commit a nation to armed conflict, the same words can send millions into battle. The words themselves have no intrinsic power. Their force comes from the social systems that recognize them.
Searle's System
Austin died in 1960, leaving his ideas in provocative but incomplete form. His student John Searle took up the project of systematizing speech act theory, publishing Speech Acts: An Essay in the Philosophy of Language in 1969.
Where Austin had catalogued examples and explored complications, Searle built rules and taxonomies. He identified five basic categories of illocutionary acts, based on what the speaker is trying to accomplish:
Assertives commit the speaker to the truth of something: "It's raining," "The meeting starts at three," "I believe she's lying." These aim to describe how the world is.
Directives attempt to get the hearer to do something: "Close the door," "Could you pass the salt?", "I order you to leave." These aim to change the world by prompting action.
Commissives commit the speaker to future action: "I promise to call you," "I'll definitely be there," "We guarantee your satisfaction." These create obligations for the speaker.
Expressives convey psychological states: "I'm so sorry," "Congratulations!", "I apologize for the confusion." These express how the speaker feels about something.
Declarations bring about changes in reality by their very utterance: "You're fired," "I now pronounce you married," "I hereby resign." These are the purest performatives, where saying makes it so.
This taxonomy isn't perfect—many utterances blur category lines or belong to multiple categories—but it provides a useful map of the territory. More importantly, Searle connected speech acts to mental states. Assertives express beliefs. Directives express desires. Commissives express intentions. This link between words and minds helped explain how speech acts work: they're the external face of internal states, governed by social conventions that make them recognizable and effective.
The Indirection Problem
Here's a puzzle that occupied speech act theorists for decades: How do we understand sentences that don't mean what they literally say?
"Can you pass the salt?"
Grammatically, this is a yes-or-no question about your ability to pass salt. A literal answer would be "Yes, I can" or "No, my arms are broken." But no one actually interprets it this way. Everyone understands it as a request to pass the salt. The question form is just packaging.
These are indirect speech acts, and they're everywhere in human communication. "Would you mind closing the window?" is not really asking about your mental state. "I was wondering if you could help me" is not a report on your cognitive activities. "It's cold in here" said by your shivering colleague is probably not a weather update.
Why do we communicate so indirectly? Largely because of politeness. Direct commands ("Pass the salt," "Close the window," "Help me") can feel abrupt, even rude. By framing requests as questions or statements, we give the other person room to decline gracefully. We soften the imposition. We maintain social harmony.
But this creates a communication puzzle. If the literal meaning points one direction and the intended meaning points another, how do hearers figure out what speakers actually mean?
Searle proposed that we work it out through inference. We hear "Can you pass the salt?" We recognize that asking about someone's ability to perform a trivial physical action would be a pointless thing to do. We assume the speaker is being cooperative—saying something relevant and useful. We conclude they must mean something beyond the literal question. We search for a plausible intended meaning. We arrive at: this is a request.
This inferential process happens instantaneously, below conscious awareness. We don't laboriously reason through the steps; we just "get it." But Searle argued that the logical structure of the inference is still there, even if we process it automatically.
Predecessors and Parallels
Austin and Searle are usually credited with founding speech act theory, but intellectual history is rarely so neat. Scholars have since uncovered a rich prehistory of similar ideas.
The eighteenth-century Scottish philosopher Thomas Reid distinguished between "solitary" mental operations (like imagining or remembering) and "social" acts of mind (like asking, commanding, and promising). Social acts, Reid argued, require expression to another person. You can remember something privately, but you can't promise something privately. A promise that no one hears isn't a promise at all. Reid was already identifying the essentially interpersonal, action-like character of certain utterances.
Even more striking is the work of Adolf Reinach, a German phenomenologist who died in World War I. In 1913—more than four decades before Austin's Harvard lectures—Reinach published a detailed analysis of what he called "social acts": promising, commanding, requesting, warning, and related performances. He argued that these acts create new normative relations between people. When you promise to do something, you create an obligation in yourself and a corresponding claim in the person you promised. These obligations don't depend on moral principles or legal enforcement; they spring into existence from the social act itself.
Reinach's work was largely forgotten for decades, overshadowed by other developments in phenomenology. Only later did scholars recognize him as having developed a sophisticated speech act theory independently of the Oxford tradition.
In linguistics, the Slovene philologist Stanislav Škrabec analyzed what he called "performative" uses of tense in the late nineteenth century. The German psychologist Karl Bühler used the term Sprechhandlung—literally "speech action"—in his 1934 work on language functions. Even Hegel, in his lectures on the philosophy of right, described speaking as a form of acting, noting that utterances can actualize and transform social reality.
The recognition that words can be deeds, it turns out, has surfaced repeatedly across different intellectual traditions. Austin gave it its most influential formulation, but he was far from the first to notice.
Beyond Philosophy
Speech act theory began as a contribution to the philosophy of language, but it has spread far beyond its origins. The framework turns out to be useful for thinking about communication in many domains.
In developmental psychology, researchers use speech act categories to track how children learn to communicate. Before children can form complete sentences, they're already performing speech acts: requesting things (by reaching and vocalizing), greeting people (by waving), protesting (by pushing away or crying). Learning language, on this view, isn't just learning vocabulary and grammar. It's learning how to do things with words—how to perform the social actions that language makes possible.
In computer science, speech act theory has influenced how we design systems that communicate. When a computer agent sends a message to another agent, what is it doing? Is it informing? Requesting? Committing? Researchers building multi-agent systems have drawn on speech act categories to model these interactions. The idea of "commitment" has been particularly useful: when an agent makes a promise, it takes on an obligation that affects how it should behave going forward. Message protocols can be designed around the creation and discharge of such commitments.
The same framework helps explain human-computer interaction. When you give a voice command to your phone, you're performing a directive. When the phone says "Setting timer for 10 minutes," it's performing a commissive. When it says "Done," it's performing an assertive that also serves as an indirect declaration—the timer really is now set. Thinking in speech act terms helps designers understand what users expect from these interactions and what kinds of responses feel appropriate.
In political science and international relations, speech act theory has influenced "securitization" theory. The idea is that certain threats become "security issues" not because of objective features of the situation, but because powerful actors successfully perform the speech act of declaring them security threats. When a government labels something a national security concern, it's not just describing a pre-existing danger; it's performing an act that changes what responses are considered legitimate. The words create the political reality they purport to describe.
In law, the performative dimension of language is impossible to ignore. Legal systems run on declarations: "guilty," "not guilty," "I do," "I sentence you," "I hereby transfer ownership." Courts and legislatures don't just describe the law; they create it through speech acts. Legal philosophers have explored how felicity conditions work in legal contexts—what gives a judge the authority to sentence, what makes a contract valid, how declarations interact with institutional power.
Economists and sociologists have extended this thinking to financial markets. When a credit rating agency declares a country's debt to be "junk," that declaration affects the country's borrowing costs, which affects its ability to pay its debts, which may turn the declaration into a self-fulfilling prophecy. Economic models, on this view, aren't just tools for predicting markets; they're speech acts that shape the behavior they claim to describe.
The Metalocutionary Level
Later theorists have added refinements to Austin's original three-part framework. One useful addition is the concept of "metalocutionary acts"—speech acts that organize or comment on the discourse itself rather than advancing its subject matter.
When a speaker says "To summarize," or "Let me put it another way," or "I'll get to that in a moment," they're performing metalocutionary acts. They're not making claims about the world; they're managing the conversation. They're signaling structure, marking transitions, directing attention.
This category helps explain aspects of spoken language that might otherwise seem puzzling. Why do speakers say "You know what I mean?" They're not really asking a question; they're checking in, maintaining connection, ensuring the discourse is working. Why do we say "Well..." at the start of certain answers? It's a signal that what follows may be more complicated than expected. These utterances don't have locutionary content in the usual sense—they're not really "about" anything—but they're doing important work in managing how communication proceeds.
Punctuation and intonation perform similar functions in writing and speech respectively. The semicolon between two clauses, the pause before a punchline, the rising tone at the end of a question—these aren't part of what's being said, exactly, but they shape how it's received. They're metalocutionary markers.
The Social Construction of Reality
Perhaps the deepest implication of speech act theory is how much of social reality depends on performative utterances. Marriage, money, property, citizenship, treaties, corporations, laws—none of these exist independent of human conventions, and all of them are created and maintained through speech acts.
Consider a twenty-dollar bill. Physically, it's just paper and ink. What makes it money is a complex web of declarations: legal tender laws, Federal Reserve policies, bank charters, all backed by speech acts of institutions with recognized authority. If everyone stopped treating those declarations as valid, the paper would be worthless. The "reality" of money is entirely performative.
Or consider a corporation. A corporation doesn't exist anywhere you can point to. It's not the building (buildings can be sold without affecting corporate identity), not the employees (they come and go), not the products (those can change entirely). A corporation exists because certain speech acts—articles of incorporation, state filings, board resolutions—bring it into being and sustain its existence. It's a creature of declarations.
John Searle himself extended speech act theory in this direction, developing what he called a theory of "social ontology." His book The Construction of Social Reality argues that human institutions are fundamentally composed of rules of the form "X counts as Y in context C." A piece of paper counts as money in the context of a modern economy. A certain building counts as a church in the context of a religious community. These "constitutive rules" create the very activities and institutions they describe, and they're implemented through speech acts.
This perspective is both empowering and unsettling. Empowering because it reveals how much of our world we create through language: if we can declare things into existence, we have more power to change social reality than we might think. Unsettling because it shows how fragile our institutions really are: they depend on continued belief in the validity of foundational speech acts, and that belief can erode.
Intentions and Actions
Kent Beck, the software engineer and author, had an illuminating customer service experience that captures something essential about speech acts. He needed to change an international flight and opened a chat with the airline. The interaction was smooth, efficient, the problem solved quickly. But Beck found himself wondering: was he talking to a human or an artificial intelligence? And did it matter?
From a speech act perspective, the question is interesting. The chat interface performed all the relevant speech acts: acknowledging Beck's request, making commitments, confirming the change. Whether those acts came from a human brain or a neural network, they had the same illocutionary force. A promise is a promise. A confirmation is a confirmation. The effects on Beck—relief, satisfaction, completed rebooking—were the same regardless of what produced the words.
But there's something unsettling about that equivalence. We tend to think that speech acts depend on intentions, and intentions seem to require a certain kind of mind. When a human customer service agent says "I'll take care of that for you," they're (presumably) expressing an intention to help. When an AI says the same thing, is there an intention behind it? Or is it mimicry—the surface form of a speech act without the underlying mental state?
Searle famously argued, in his "Chinese Room" thought experiment, that computers could manipulate symbols without understanding their meaning. If he's right, then computer-generated speech acts might be zombies: they look like the real thing, they function like the real thing, but something essential is missing.
Then again, maybe what's "missing" doesn't matter as much as we think. The felicity conditions for most speech acts are about external circumstances, not internal states: does the speaker have authority? Is the procedure correct? Are the circumstances appropriate? When a court's computer system generates a form that declares a judgment satisfied, it's performing a declaration—and the legal reality it creates is just as real as if a human had typed it.
As artificial intelligence becomes more capable, these questions will become more pressing. But they were already implicit in speech act theory from the beginning. The theory always emphasized that the power of words comes from social conventions, not from any intrinsic property of the speaker's mind. In a sense, all speech acts are socially constructed—including the speech acts of humans.
The Performing of Everyday Life
Once you start noticing speech acts, you see them everywhere.
When your friend says "I'm sorry," they're performing an expressive. Whether it "works" as an apology depends on felicity conditions: Do they seem sincere? Was there something to apologize for? Do you accept it?
When your boss says "Let's touch base Friday," they're performing a directive—mildly phrased, but a directive nonetheless. The meeting is not really optional.
When you say "I'll be there at seven," you're performing a commissive. You've created an obligation. If you don't show up, you'll have violated something—not a law, not even an explicit promise, but the commitment your utterance created.
When a website says "By clicking 'Agree,' you accept our terms of service," it's invoking a performative framework. Your click is meant to count as a declaration of agreement. Whether it really does—whether clicking constitutes genuine consent—is a live debate in law and ethics.
Even silence can be a speech act. In certain contexts, saying nothing counts as assent. In others, it counts as refusal. The meaning of silence, like the meaning of words, depends entirely on social conventions and expectations.
This is perhaps the most important lesson of speech act theory: language is not a window through which we view the world, but a tool with which we build it. Every utterance is an action, with consequences that ripple outward through social reality. We are all, in every conversation, doing things with words.