Grice's Logic and Conversation: Conversational Implicature
In his 1975 essay “Logic and Conversation,” H. P. Grice explains how people work out conversational implicatures through context and the Cooperative Principle.
PhilosophyAll chapters Reading tools Notes & comments
What speakers imply › What a short recommendation can conveyFull chapter · Sections in this chapter1 / 8 ↓
What a short recommendation can convey
A professor is asked to recommend a former student for a philosophy position. The letter singles out two things. The student's English is excellent and they attended tutorials regularly. If the committee expects an assessment of the student's work in philosophy, the omission will stand out. A committee member might conclude that the professor thinks poorly of the work, even though the professor never said so. H. P. Grice uses a recommendation of this kind to show how people can communicate by giving an audience good grounds for an inference. Grice, “Logic and Conversation,” pp. 51–53.
Change the setting and the letter can communicate something different. If the form asks only about English and attendance, the same two sentences can be a complete answer. If the writer taught the student long ago and mentions that they have never read the student's recent work, the omission can indicate limited knowledge. What the sentences say can survive these changes while what they imply changes with the occasion. The challenge Grice takes up is to explain how a reader ordinarily gets from one to the other.
What is said and what is implicated
Grice connects what is said closely with the conventional meaning of the words. Even when a speaker is speaking literally, establishing precisely what they said can require finding out whom a pronoun refers to, when the utterance was made, or which sense of an ambiguous word is at issue. In the recommendation the reader needs to know which student the writer is referring to. The reader also needs to know what kind of assessment was requested to work out why the letter is so brief. The two uses of context help answer different questions. Grice, pp. 43–45.
Grice begins the essay with a dispute about pairs such as ordinary “and” and its counterpart in symbolic logic. One tempting explanation for the difference is that the ordinary word has additional meanings that a logical connective leaves out. In suitable circumstances, “Sam locked the door and turned off the light” suggests that Sam did those things in that order. There are also circumstances in which the order is irrelevant. If someone asks which of two assigned jobs Sam finished, the natural answer may use “and” in either order. Grice offers an account of how some apparent differences between logical devices and ordinary words can result from what it is reasonable to infer in a particular exchange. The essay does not settle every question about temporal connectives. Grice, pp. 41–43.
Not every implicature in Grice's account is conversational. A particular expression can carry an implication by convention. In the paper's own example, a speaker says someone is English and uses “therefore” to link that claim to the claim that the person is brave. Grice treats the speaker as saying that the person is English and brave, while the connective conveys a relation of consequence. The inference in the recommendation works differently. Its words would be entirely apt in an exchange about attendance and fluency. An implication of poor philosophical ability arises from the fit between the short reply and the assessment the committee asked for. Grice, pp. 44–45.
The Cooperative Principle and the maxims
Grice describes conversation as an activity in which the participants can generally recognize a purpose or direction. On his first approximation of the Cooperative Principle, speakers should contribute what is required by the accepted direction of the exchange at that stage. That direction can change. A request for the location of a building may turn into a request for the fastest route there. In an everyday conversation its direction can be looser still. Agreement on the object of the exchange gives its participants a basis for expecting intelligible contributions even if they disagree about the answer. Grice, pp. 45–46.
Consider two travelers deciding whether they have time to catch a bus. They ask a person nearby when it leaves. The departure time is more helpful than a list of the buses on that route for the entire week. Guessing that the bus leaves in ten minutes could sound helpful but give the travelers poor grounds to rely on the answer. A long description of a different route will mean little unless there is some reason the travelers should consider it. Even the right time is hard to use if it is delivered as an ambiguous reference to a stop with two departure boards. Grice separates those aspects of an answer:
- Quantity concerns how much information the exchange calls for. Grice also proposes avoiding unnecessary information and notes that some of its effects might be explained through relevance.
- Quality concerns truthfulness and evidential support. Speakers are expected to avoid claims they believe false or have inadequate reason to make.
- Relation asks whether the contribution is relevant to the exchange as it stands.
- Manner concerns how the contribution is made. This includes avoiding needless obscurity and ambiguity and arranging what is said intelligibly.
The labels divide questions the hearer can ask about one contribution. They need not correspond to separate steps taken by the people talking. Grice also considers analogies with coordinated action. A partner helping you make a cake has reason to offer the ingredient you need at that point and to give you a usable amount. Conversation can require a similar judgment about what might help someone who is pursuing a common activity. Grice, pp. 45–48.
How the inference works
One of Grice's speakers tells another that their car is out of petrol. The reply mentions a garage around the corner. If the second speaker thinks the garage might be open and sell petrol, the reply is a useful contribution. If they know it has been boarded up for years, the same remark would puzzle a driver unless the speaker signals a different purpose. The driver makes sense of the remark by recognizing what the speaker must at least be treating as possible. The garage need not turn out to be open. The point concerns what the speaker led the driver to believe was a live possibility. Grice, pp. 50–51.
The inference starts from what was said, relevant background, and a reason to see the contribution as serving the exchange. The driver knows that the other person knows why a garage matters. The speaker may expect this to be enough for the driver to recognize the point of mentioning it. An explanation of the inference can set these facts out explicitly even when the driver understands it immediately. Grice requires that a conversational implicature in principle be capable of being worked out in this way. His account is a reconstruction of the inference, without taking a position on how fast or consciously the listener reaches it. Grice, pp. 49–51.
The garage remark is relevant because the supposed availability of petrol connects it to what the driver said. No maxim has to be openly flouted for this implicature to arise. The case also shows why the outcome can be weaker than a guarantee that the garage is open.
A listener can draw conclusions unrelated to what the speaker meant to communicate. If a driver hears the speaker's accent and guesses where they grew up, that guess is not on its own an implicature of the remark about the garage. The conversational sort of inference asks why the speaker made this contribution and what they can reasonably count on the hearer to recognize. The reference letter conveys an unfavorable judgment only when the reader has a basis for thinking its limited praise was left to do that work. A writer who filled in the wrong form by accident has created a misleading impression without necessarily communicating what the committee infers.
When a maxim is left unfulfilled
Grice separates four ways in which a maxim can fail to be met. A speaker can quietly violate it and risk misleading the audience. They can explicitly opt out, as when declining to answer a question. They can face a clash between two expectations. Finally, they can flout a maxim by making their departure obvious so that the audience will look for a further point. A single brief reply can play any of those roles. To judge which it plays, a reader has to know why the speaker gave only that information and what they expected the listener to be able to infer. Grice, pp. 48–49.
Take Grice's example of a traveler asking where a friend lives so that they can decide whether to visit. The reply places the friend somewhere in the south of France. The location is too broad to plan a visit. If the speaker could only give a more precise answer by guessing, quality and quantity pull in different directions. The limited answer can convey a limit on what the speaker knows. Someone who says “I can't disclose the address” is doing something else. They make their refusal explicit and leave open how much they know. The quantity of useful information may be similar in the two cases even though its shortage has a different explanation. Grice, pp. 49 and 51–52.
The short recommendation makes sense as a flouting case when the writer could reasonably be expected to offer a substantive appraisal and the lack of it is obvious. The departure from quantity leads the committee to ask what the writer hoped they would notice. Quality can also be flouted. Suppose a friend betrayed a confidence and everyone present knows it. Another person pointedly praises the friend's loyalty. The audience can recognize that the speaker is making as if to assert something they do not believe and can look for the criticism the remark conveys. Grice presents a distinct comparison for metaphor. Someone who compares a person to an ingredient in their coffee can invite a search for a relevant resemblance. The mere denial of the literal comparison would add almost nothing. Grice, pp. 52–54.
Some apparent departures are simply miscues. A speaker answers the wrong question because they arrived halfway through a discussion. The answer may come across as evasive, but its irrelevance does not necessarily signal a further communicative intention. A theory that calls every irrelevant contribution a flout loses the difference between a remark designed to be understood indirectly and a remark that landed in the wrong exchange.
Cancellation and the way something is phrased
Imagine a teacher telling another teacher that some students passed the test. In a conversation about whether everyone passed, that may suggest that some failed. The hearer assumes the teacher knows whether everyone passed and would ordinarily give the stronger answer if it were available. If the teacher is reading the first stack of papers and has not seen the rest, those assumptions lapse. The very same words can then be an accurate update without saying how the rest did.
The speaker can also add “in fact, every student did” without contradicting the initial statement that some passed. This illustrates explicit cancellation. The circumstances of the stack of papers illustrate contextual cancellation. Compare trying to follow the statement that some passed with “none of them passed.” That would conflict with the stated content. Cancellation helps the hearer determine what a sentence requires and what was supplied by assumptions about the purpose of the exchange. Grice discusses the two sorts of cancellation in connection with generalized conversational implicature. The test is evidence about a case and should be used alongside the reasoning that generated the inference in the first place. Grice, pp. 56–58.
Grice also distinguishes implicatures that depend closely on a particular setting from those that are normally conveyed by a form of words across many settings, unless there is information to override them. The garage example needs the driver and the broken-down car. A familiar turn of phrase can invite an inference even before much has been said about the exchange. The fact that an inference comes readily to mind in the second case does not establish that it is conventionally encoded. It can still disappear when the occasion gives the hearer a reason to read the phrase differently. Grice, pp. 56–58.
Wording makes another difference. If an inference rests on the content of a remark plus the shared purpose, saying the same thing in other words often leaves the inference intact. Grice calls this non-detachability and qualifies it when the manner of expression matters. Suppose someone describes a familiar action using conspicuously elaborate language when a plain phrase would have done. That phrasing might imply that the action was unusual in some way. A shorter expression with the same content need not invite that reading. Cases like these help locate whether an implication depends on what was said, the way it was said, or the use of a particular conventionally loaded word. Grice, pp. 55 and 57–58.
An implication can also be uncertain in scope. If two visitors are invited to a picnic and a friend says one of them dislikes crowds, the friend might be suggesting where the group should sit, why the guest may arrive late, or why the guest was hesitant about coming. The inquiry depends on who was discussing the seating and which concern prompted the remark. Grice does not require a sentence to carry a single sharply specified conversational implication in every use. Several explanations for its contribution can remain viable. The indeterminacy is useful evidence in a particular case. A listener who needs to know whether they should move the picnic can ask about that implication directly. Grice, pp. 57–58.
What the account leaves to the listener
Listeners can disagree about what would have been an adequate answer. Suppose a professor highlights English and attendance believing those are precisely the matters the search committee wants assessed. The committee might interpret the brevity as faint praise if it expects every writer to comment on scholarly ability. To decide whether the professor actually implicated a negative judgment, you would need to know which expectations were shared or could reasonably have been taken to be shared. An audience can also infer too much when several interpretations would make sense. Grice recognizes that more than one account of the point of an utterance can remain open. Grice, pp. 50 and 57–58.
The uncertainty is much smaller in his example of the garage because the speaker and driver know what problem prompted the exchange. If the driver asks “Can I buy petrol there now?”, the follow-up can clarify what the speaker took their own answer to be contributing. The usefulness of that follow-up depends on distinctions the original reply left open, such as whether the speaker knows the place is open or merely thinks there is a chance. Grice's idea turns a seemingly immediate understanding into a question you can investigate. You can ask which inference a speaker was relying on, what made it available, and what change of circumstances would withdraw it.
Further reading
- H. P. Grice, “Logic and Conversation” (1975), pp. 41–58. Pages 43–50 lay out the distinction between saying and implicating and the maxims. The examples and the discussion of cancellation and detachability begin on pp. 51 and 56.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!