What is a conversation? A simple question that's difficult to answer. We think we know until we try to explain it to someone else.
In The Domestication of Language, Daniel Cloud explores the wonderful world of conversations.
Our difficulty in accounting for conversation isn’t a sign that nobody’s ever tried to understand it. The intense focus on rhetoric by classical philosophers, for example, was the organized study of a certain rather formal kind of public conversation, and our interest in the phenomenon has continued until the present.
Cloud points us to H.P. Grice and his 1975 paper “Logic and Conversation.” Grice, according to Cloud, argues that human conversations happen around a shared purpose.
We sometimes may be deluded in thinking that such a shared purpose exists—for example, when talking to a confidence man—but the supposition is required to make us willing to participate. The purpose may be obvious—the car is out of gas, we have to figure out what to do—or frivolous, extremely serious, or horrific. The torturer seeks to create a common interest so he can have a truthful conversation with us, even though his method involves the stick and not the carrot.
Grice admitted that he was perplexed about the exact nature of the understanding involved. … Common interests in the absence of enforceable contracts create coordination games … Each of us would rather converse on some mutually agreed topic than not converse at all, provided that all the others do. It isn’t a helpful form of participation in the conversation to periodically interject irrelevant remarks on completely unrelated topics, so we would prefer that all participants converse about the same topic as everyone else is or at least change the subject in culturally acceptable, legitimate ways. We often would be happy to converse about some other, slightly different topic if that topic had been raised by one of the participants instead. There always are alternatives, unless the people are enacting a play or another ritual, and real conversations change and drift as they go along, so the topic may well morph into one of those almost equally good alternatives. The conversation may acrimoniously disintegrate into no conversation, on no topic, if it goes badly, or it may gently evaporate into a resolve to have other conversations later. There always are different conversations we could have had instead. If someone new enters the discussion, we’d prefer that he stick to the topic, though if we were discussing something else, we’d prefer that he discuss that instead.
Topics are temporary and convey and establish conventions. They are also malleable and complex, changing directions in real time.
A shared common ground is first established, and then it’s extended and amended by the successive remarks of those involved. The changes may be incremental, or—if it’s possible to bring the other participants along with us, if people are agreeable and the transition isn’t too complicated to be made in unison without much preparation—they may be abrupt.
Conversations are not intended for everyone. From professions that have specialized vocabulary to conversations so generic only people who know each other would understand the hidden meaning, we have ways of excluding people, who, even if physically present, will not be a participant or even benefit.
What’s also true of most conversations is that not just anyone can participate. Perhaps we all haven’t been properly introduced. Or the conversation may be one that only topologists or elk hunters or members of the president’s national security team can engage in, or one that only Romeo and Juliet can be a part of. We may seek admission to a conversation and be welcomed or rebuffed. Yet this isn’t usually because there’s something scarce being shared by those conversing, which they would necessarily receive less of if someone else participated. Although there are conversations like that, many conversations are not. Sometimes new participants, even excluded ones, would have added something. In the language of economics, conversations are excludable and non-rivalrous. People can be prevented from benefiting from them or they can be excluded, but those who share in them don’t necessarily diminish their worth for the others. …
It seems that a conversation—like the highway system or the community that speaks Welsh—is a particularly informal, spontaneous, and fleeting club, an ephemeral microinstitution that flickers into and out of existence in a few seconds, minutes, hours, days, weeks, or years after its initial convening and that is organized around a temporary set of conventions about its topic, manner, and so on. By seeking admission, we represent ourselves as willing to conform to these conventions unless we can persuade the other participants to amend them. Sometimes some of the conventions established in a conversation also acquire contractual force—for example, when the conversation itself is a negotiation—but many do not.
How can participants in a conversation advance a common interest? Grice argues that most of what is conveyed in conversations is “implicature.”
Consider the following exchange:
a: Will Susan be at the game?
b: She has to teach that day.
In Grice’s terminology, B has implicated, but not said, that she won’t be at the game. … This conclusion depends on the common knowledge, known by both participants to be known by both participants, that teaching would preclude going to the game, perhaps because they will take place at the same time. Knowing this, A can work out what B is trying to tell him, what B is attempting to implicate
Grice distinguishes between this sort of context-dependent, situational implicature, which he calls “conversational implicature,” and mere conventional elisions of the following kind: “Socrates is a man, and therefore mortal.” Here I’ve left out a premise that would be required for the “therefore”; I’ve neglected to mention that all men are mortal. But I didn’t have to, because you and I, like everyone else, already know that. Without having to think about it, you naturally will assume that I am assuming that you will extract this information from the incomplete argument I’ve offered. Grice calls this slightly different phenomenon … “conventional implicature.”
How do we work out the intended conversational implicatures? Through various maxims.
[T]hese maxims (are organized) under the more general principles of quantity, quality, relation, and manner. We assume that the speaker is telling us as much as we need to know for the purposes of the conversation, but no more (quantity). We assume that he’s attempting to tell us only things that he knows to be true and is not asserting things that he believes to be false or for which he has no evidence (quality). We assume that what he’s saying is somehow relevant to the mutually understood, though constantly evolving, topic of the conversation (relation). We assume that he’s attempting to be perspicuous, that he would prefer to avoid ambiguity and obscurity, avoid prolixity, and present his narration or his argument in an orderly way (manner).
In answering A’s question about Susan, B must be understood to be telling A as much as he needs to know for his question to be answered. Likewise, A must assume that B believes it to be true that she has to work and has reasonably good grounds for that belief. A must assume that this information is somehow relevant to the topic raised by his question. Assuming these things, A is in a position to interpret B’s remark as intended to produce the implicature that Susan will not be at the game because it conflicts with her work. If her work has a special relationship to the game or its venue that means that the remark should produce the opposite conclusion, then B has failed to follow the principle of quantity correctly, because he’s left out something he would have had to tell A to make his remark interpretable. He has assumed the existence of a piece of common ground that’s actually missing.
Common knowledge is really the key to conversations because it's the key to common ground.
Common knowledge, first created as common ground in formal or informal conversation and then conserved and referred to in later conversations, marks the boundaries of skill-centered speech communities, of the subcommunity of shamans or eel farmers or navigators or structural biochemists or Shinto priests. These are things that these people must know in order to converse with one another, making them unable to converse as freely with people who lack their skill set.
In conversation, the method used for creating new items of common knowledge is the participants explicitly or implicitly informing one another of things, so conversations create parts of their own common ground as they go along. By so doing, the participants may become partly isolated from the rest of their speech community, which now doesn’t share the newly created common ground. New tacit conventions also are negotiated indirectly and obliquely in particular conversations, by means of concerted choices among competing, unstated alternatives, which can make it even harder for an outsider to follow them.
A conversation consists of a series of its participants’ dovetailed and cumulative modifications of their common ground, and at the end of the conversation, they may share different knowledge or intentions (“Then yes, let’s do that”) or expectations (“Well, then I guess we can expect the same thing to happen every time”) or different explicit conventions (“OK, I guess next time, whoever called originally should be the one to call back”) than they did before.
This new knowledge has become common knowledge in the group conversing and now can be used as such, can be assumed to be part of the common ground for subsequent discussions by the same group. B will expect A to remember that Susan has to work on the day of the game. We’ll be expected to remember the new plan or expectation or convention that’s finally been arrived at.
Cloud argues that conversations convert knowledge, expectations and beliefs from private knowledge to common knowledge within the conversing group.
What is common knowledge can support conventional (as opposed to conversational) implicatures, so the group’s stock of possible conventional implicatures is enlarged as a result. From now on, it may not be necessary to mention that Susan has to work on the day of the game; perhaps it can simply be assumed. Every successive conversation among a certain group means that less must be said in subsequent conversations, that more and more can be “taken for granted.”
This fact can produce a sort of cultural microversion of songbirds’ local dialects, local, group-specific assumptions that make it harder and harder for newcomers who lack the same shared history to participate in the group’s conversations. Conversations make us clannish; they erode the barriers to communication and trust within the group while erecting new ones around it, in a tiny, temporary, ultrafast cultural version of one of John Maynard Smith and Eörs Szathmáry’s (1998) “major transitions.” A conversation creates a club that subsequently may function in some ways like a single, self-interested unit, which may see itself as competing with other, rival clubs and may exclude interlopers or impose its own rules on new entrants.
Conversations are much wider than the term implies. In fact, no words need be spoken at all.
When the master holds out his hand for a hammer, the apprentice can understand the gesture as a request of that kind only because he assumes that the master isn’t making an unnecessary gesture, isn’t trying to trick him, is asking for something relevant to the collaborative task at hand and not his hunting spear, and isn’t making a gesture he thinks the apprentice will be unable to interpret.
If you're into the evolution of language, you'll love The Domestication of Language.