Implicature: Saying Without Saying

Learning outcomes

Most of what people mean is never actually said. A teammate writes “Some of the tests pass,” and you are expected to hear “not all of them” without being told. This page is about that gap between the words and the message, and about the quiet rules that let a listener cross it reliably. Once you can see the gap, a great deal of polite, indirect, and ironic English stops being confusing and starts being readable.

After studying this page, you can:

• Explain what an implicature is and why a listener can work out a meaning the speaker never stated.
• State Grice’s cooperative principle and its four maxims in plain language, and show how each one generates an implicature.
• Tell apart a maxim quietly observed from a maxim deliberately flouted, and read the extra meaning a flout carries.
• Recognize indirect speech acts, where a question or a comment is really a request or a hint.
• Read an implicature by asking what a cooperative speaker would mean here, and know when not to over-read.

Before we dive in

Here is a short exchange. A manager asks, “Is the deploy done?” An engineer answers, “The CI is still running.” Notice that the engineer never said “no.” Strictly, the answer is about continuous integration, not about the deploy at all. Yet the manager hears “not yet, wait” instantly and without effort. Nobody is confused. Nobody asks for clarification.

This is strange if you think language works by literal meaning alone. The literal content of “The CI is still running” is just a fact about a process. The message the manager actually receives, that the deploy is not finished, is nowhere in those words. Something is letting the manager add a meaning the engineer did not state.

That something is the subject of this page. The extra meaning a listener works out, beyond the literal words, is called an implicature, and the act of conveying it is to implicate (not to imply in the loose everyday sense, though they overlap). Implicature is not rare or literary. It is the ordinary machinery of conversation, running under almost everything people say. The philosopher H. P. Grice gave us the first clear account of how it works, and that account is where we start.

Breaking it down

1. The cooperative deal that makes inference possible

Start with the puzzle from the opening: how can the manager be sure that “The CI is still running” means “not yet”? The engineer could, in principle, have been changing the subject, or stating a random fact. The manager does not consider those readings for a moment. Why not?

Grice’s answer is that conversation rests on a shared, unspoken agreement. When two people talk, each assumes the other is trying to communicate, to contribute something useful to the shared goal of the exchange. Grice called this the cooperative principle: make your contribution what is required, when it is required, by the direction of the talk you are in. You assume your partner is cooperating, and they assume the same of you.

This assumption is the engine. Because the manager assumes the engineer is cooperating, the manager reasons roughly like this: “She is a sensible person answering my yes-or-no question. If she mentions the CI instead of saying yes, she must be telling me something relevant to my question. The relevant thing is that the deploy depends on the CI, which is not finished. So she means: not yet.” The implicature is not in the words. It is reconstructed by the listener, on the assumption that the speaker is playing the cooperative game.

2. The four maxims, and how breaking them quietly creates meaning

The cooperative principle is broad, so Grice broke it into four practical guidelines, called maxims. They describe what a cooperative speaker is expected to do. The point is not that people always obey them. The point is that listeners assume they are being obeyed, and that assumption is exactly what lets a listener calculate an implicature.

Read the diagram from the top. The cooperative principle sits above four maxims, and each maxim, when a listener assumes it is in force, generates a specific kind of inference. Take quantity first, because it is the most common. A cooperative speaker gives as much information as the situation needs and no more. So if a colleague says “Some of the tests pass,” you assume they gave you the full relevant picture. If all the tests had passed, the cooperative thing would have been to say “all” or “they pass.” Choosing the weaker word “some” signals that the stronger claim does not hold. You conclude: not all pass. That conclusion is a quantity implicature, and it is entirely the listener’s work.

The maxim of relation, or relevance, drives the deploy example: you assume the answer bears on your question, so you connect the CI to the deploy yourself. Quality (be truthful) and manner (be clear and orderly) round out the set. Each maxim is a default the listener trusts, and that trust is what converts plain words into richer messages.

The table below collects several everyday utterances so you can practise the move from literal meaning to implicature, and name the maxim doing the work each time.

Work across one row to feel the pattern. The literal column is thin and often almost empty of useful content. The implicature column is what the speaker actually wanted to land. The gap between the two columns is the implicature, and the last column names the maxim that lets a cooperative listener cross it. Notice that “I did not break the build” implicates that someone else did, by quantity: a fully cooperative, fully informed speaker who knew nothing was broken would not single out their own innocence, so mentioning it suggests there is something to be innocent of.

3. Flouting: breaking a maxim on purpose so the listener reads more

So far the maxims have been quietly observed, and the implicature comes from trusting them. Now comes the sharper move. A speaker can break a maxim openly, in a way the listener is meant to notice, precisely so the listener reasons, “She is still cooperating, so she must be breaking this rule to tell me something extra.” Grice called this flouting a maxim.

Suppose a colleague proposes a clumsy fix, and you reply, “That is certainly a bold choice.” On the surface this praises. But you have flagrantly said too little for the situation (a flout of quantity) and arguably praised what does not deserve it (a flout of quality). Your listener, still assuming you cooperate, looks for the meaning behind the gap and lands on the real message: you do not approve. This is the polite art of damning with faint praise. “He is technically competent” works the same way: it is true, but it is so thin, so carefully limited, that the listener infers the things you pointedly did not say, that he is hard to work with, perhaps, or uninspired.

Flouting is also the engine of irony and sarcasm. When someone says “Great, another meeting” about an unwelcome invitation, they openly say the opposite of what they believe, a flagrant flout of quality, and trust you to flip it. The mechanism is identical to the polite flout above. Because this is so central to humor, it has its own page: see how humor, irony, and sarcasm work.

4. Indirect speech acts: when a question is really a request

There is a closely related pattern where the literal form of a sentence and its real job come apart. “Can you pass the salt?” is shaped like a question about your arm’s ability. Nobody treats it that way. It is a request, and you answer it by passing the salt, not by saying “Yes, I can.” This is an indirect speech act: one kind of sentence (a question) used to perform another kind of act (a request).

The link to implicature is direct. A literal yes-or-no answer would be unhelpful and slightly absurd, which signals that the literal reading is not the intended one. Assuming cooperation, you look for the act the speaker actually wants performed. The interrogative form is doing politeness work: asking rather than ordering softens the request, a theme developed in politeness, face, and indirectness.

The same move covers a wide range of everyday English. “It is getting late” is shaped like a remark about the time but often works as a proposal to leave. “Have you considered the cache?” is shaped like a question but is usually a suggestion that you should look at the cache. “Would you mind reviewing this today?” is a request wearing a question’s clothes. In each case the surface form is one thing and the intended act is another, and the listener bridges the two by assuming the speaker is contributing something useful.

If you chose the second option, you read the implicature correctly. The reviewer’s words are a question, but the act is a request, and you recover it by asking what a cooperative reviewer would mean by raising that exact point at that exact moment.

5. How to read an implicature

You now have the parts, so here is the single habit that ties them together. When the literal words feel oddly thin, weak, off-topic, or too obvious to be worth saying, do not stop at them. Ask one question: what would a cooperative speaker mean by saying exactly this, here, now? The answer is usually the implicature.

Run the habit on a fresh case. Your manager reviews your proposal and says, “Well, the formatting is very clean.” The words praise the formatting. But formatting is a small thing to single out about a proposal, so the response is suspiciously thin (a quantity signal). A cooperative manager who liked the substance would mention the substance. Singling out the formatting implicates that the substance has problems. The reading is not paranoid; it is what the cooperative assumption licenses. The skill is not magic. It is the disciplined application of the cooperative principle, asking why this and not more.

Mental Model: the literal layer and the meant layer

If one wrong picture causes most trouble here, it is the idea that meaning lives in the words, full stop. On that picture, to understand a sentence you decode its literal content and you are done. This is the model a careful non-native learner often arrives with, because it is exactly how vocabulary and grammar are taught, and it handles textbook sentences perfectly.

The better picture has two layers. The bottom layer is the literal meaning, what the words say. The top layer is the meant meaning, what the speaker intends you to take away, which is the literal meaning plus every implicature the situation generates. Fluent listeners read both layers at once and usually act on the top one. The deploy answer, the soft no, the salt request: in each, the bottom layer is nearly empty and the top layer carries the whole message.

Switching to the two-layer model is the central shift of this page. It reframes listening as a two-step act: decode the words, then ask what a cooperative speaker means by choosing those particular words. The first step you already do. The second step is the one this whole domain is training.

Where it bends: cancellability and culture

Implicature is powerful, but it is not as solid as literal meaning, and three boundaries keep you from over-trusting it.

First, implicatures are cancellable. A speaker can add words that take the implied meaning back, with no contradiction. “Some of the tests pass, in fact all of them do” is not a contradiction, because “not all” was only implied, never stated. This is also why implicature is so useful for plausible deniability: a person can implicate something and later say, with a straight face, “I never said that.” The literal words protect them. A boss who says “I did not say you had to stay late” is leaning on exactly this gap.

Second, how much is implied varies by culture and context. Some cultures and some workplaces lean heavily on implicature, leaving a great deal unsaid and expecting the listener to fill it in. Others say things much more directly, and reading a hidden message into a plain statement there is itself a mistake. The amount of meaning you should expect to be unstated is not fixed; it is a property of the people and the setting. This variation is the heart of directness across cultures.

Third, and following from the first two, do not over-read. Not every utterance carries a hidden message. “The CI is still running” might, in a relaxed setting, just be a passing observation. Treating every plain sentence as a coded signal is its own failure, and it makes you tiring and slightly paranoid to talk to. The cooperative assumption that licenses implicature also bounds it: you infer only what a cooperative speaker would plausibly mean, not whatever you can imagine them meaning.

Common mistakes

The first mistake is being too literal and missing the message. This is the classic non-native trap: you answer “Yes, I can” to “Can you pass the salt?”, or you hear “interesting idea” as approval. The words are processed; the implicature is not. The fix is the habit from section five: when words feel thin or oddly chosen, ask what a cooperative speaker means by them, and act on that.

The second mistake is the mirror image: creating an unwanted implicature without meaning to. If you say “Some of the deliverables are done” when in fact all of them are, your listener will hear “not all,” because the quantity maxim is automatic. You have implied something false by understating. The same goes for praise: say “It is technically fine” about work you genuinely admire, and you have damned it by faint praise you did not intend. The fix is to monitor your own words for the implicatures they trigger, and to say “all” when you mean all, and to praise fully when you mean to praise.

The third mistake is assuming an implicature was meant when it was not, the over-reading from the previous section. You decide that a neutral, factual reply was a pointed criticism, and you respond to a slight that was never made. The fix is to weigh whether a cooperative speaker in this setting would actually have meant the hidden message, and, when it matters, to check rather than assume.

Mastery Questions