For example, we now all realize that the speaker who says "It's me" is not violating a rule of English by which he should say, "It's I." Rather, the mistake belongs to the grammarian who calls it an error.
Speaker transition without gap or overlap is a feature of the social organization of conversation, achieved always then and there. For example, participants do not retrospectively attain it by editing their memory of a conversation. They do not, in the first instance, go outside the conversation in order to report violations to referees, policemen, oracles, etc., in the hope that external agencies will punish the violators. There is, then, a social organization to turn-taking which has as one of its proper products that one person talks at a time: Achieving this product requires participants to encounter and solve at least two tasks: the collaborative location of transition points, and the collaborative use of means for arriving at who speaks after any current speaker (Beattie, 1983). These are tasks which, on the situated occasions of their solution, are tasks of understanding. And participants so interpret them. They take failing to talk when one has been selected to and another stops as evidence of failing to understand what has been said.
The specific kinds of understanding required for achieving proper turn-taking are determined by how turn-taking is socially organized. For example, if conversation were structured so that the order of speakers and the lengths of their utterances were pre-assigned for whole conversations, turn-taking would impose rather minimal tasks of understanding upon participants (Duncan, 1972). They work in such a way as to require that parties to a conversation do extensive work of understanding if their system of turn-taking is to operate as it does. Both employ utterance units which need to be constantly monitored for their completion. Both operate to select future speakers in an one utterance at a time fashion. And both, thereby, impose upon conversation participants demanding and identical tasks of understanding and of demonstrating understanding.
If speaker transition is to occur with neither gap nor overlap, any intended next speaker must work on understanding the current utterance so as to know what it will take for that utterance to be completed. Utterances must be built so that attention to them permits projecting their future (Duncan, 1974). Participants must be trained in an ability to understand that permits them to use such information in a timely fashion. It is possible to design a system in which the work of such understanding would be minimal. A linguistic particle or other sign might be required to be emitted at some specific point before the ending of an utterance. Or utterances might have to be of some pre-specified size. But, in fact, utterance completion does not operate
In addition to the collaborative locating of utterance endings, any conversational system in which turns are taken one at a time, must have ways of allocating future speakership. Some formula which pre-assigned the order of all speakerships for a conversation might both reduce the chances of more than one speakership happening at once and assure that someone would be responsible to talk upon each completion. But the turn-taking systems of Thai and American conversation do not operate this way. Instead, they work one utterance at