This paper presents a model for the organization of turn-taking in conversation, which is examined against a list of observable facts about conversation. The model is characterized as locally managed, party-administered, interactionally controlled, and sensitive to recipient design. It is contrasted with turn-taking systems in other speech-exchange systems. The model is based on research using audio recordings of naturally occurring conversations. The model is described in terms of two components and a set of rules: the turn-constructional component and the turn-allocation component. The turn-constructional component involves the use of various unit-types, such as sentential, clausal, phrasal, and lexical constructions. The turn-allocation component involves two types of techniques: those in which the current speaker selects the next speaker, and those in which the next speaker self-selects. The rules governing turn-taking provide for the allocation of a next turn to one party and coordinate transfer so as to minimize gap and overlap. The rules also provide for the possibility of repair mechanisms for dealing with turn-taking errors and violations. The model is shown to account for a range of observable facts about conversation, including speaker-change, one party talking at a time, brief overlaps, transitions with no gap and no overlap, varying turn order and size, and varying numbers of parties. The model is also shown to be compatible with a wide range of conversational contexts and to be sensitive to the characteristics of the participants and the situation. The model is considered to be a basic form of organization for conversation, which is invariant to the parties involved and can be selectively and locally affected by social aspects of context.This paper presents a model for the organization of turn-taking in conversation, which is examined against a list of observable facts about conversation. The model is characterized as locally managed, party-administered, interactionally controlled, and sensitive to recipient design. It is contrasted with turn-taking systems in other speech-exchange systems. The model is based on research using audio recordings of naturally occurring conversations. The model is described in terms of two components and a set of rules: the turn-constructional component and the turn-allocation component. The turn-constructional component involves the use of various unit-types, such as sentential, clausal, phrasal, and lexical constructions. The turn-allocation component involves two types of techniques: those in which the current speaker selects the next speaker, and those in which the next speaker self-selects. The rules governing turn-taking provide for the allocation of a next turn to one party and coordinate transfer so as to minimize gap and overlap. The rules also provide for the possibility of repair mechanisms for dealing with turn-taking errors and violations. The model is shown to account for a range of observable facts about conversation, including speaker-change, one party talking at a time, brief overlaps, transitions with no gap and no overlap, varying turn order and size, and varying numbers of parties. The model is also shown to be compatible with a wide range of conversational contexts and to be sensitive to the characteristics of the participants and the situation. The model is considered to be a basic form of organization for conversation, which is invariant to the parties involved and can be selectively and locally affected by social aspects of context.