How a meeting works out who speaks next

A queue with no visible order

A discussion among several people requires one person to speak at a time, and somebody has to decide who. In a chaired meeting part of that job is formal, but the overwhelming majority of transitions are handled by the participants themselves, in the fraction of a second at the end of an utterance, without anybody appearing to manage anything.

The mechanism is remarkably efficient in person and remarkably fragile everywhere else. Understanding what it depends on explains most of what goes wrong in meetings, including several things routinely attributed to the character of the people in them.

Endings are signalled before they arrive

A speaker approaching the end of a point signals it in advance through several channels at once: the sentence completes grammatically, the pitch falls, the pace slows, the gaze moves back to the group. These cues arrive slightly before the actual end, which is what gives the next speaker time to prepare.

That advance warning is why fluent conversation has almost no gaps and almost no overlaps. It also explains why the ends of sentences are so heavily defended. A speaker who wants to continue keeps their sentence grammatically open, holds the pitch up, and avoids looking at anybody, which reads as thinking aloud and is functionally a request not to be interrupted.

Selecting the next speaker, or leaving it open

A turn ends in one of two ways. The speaker names or looks at somebody, which allocates the next turn directly, or they release the floor to the room, in which case whoever moves first takes it. The first produces orderly discussion and the second produces competition.

Chairs who are good at this use direct selection far more than they use general invitations, because an open question in a group of eight is a race with no starter. It is also why the quietest person in a meeting frequently speaks well when asked a direct question, having been prevented from contributing less by reticence than by an allocation mechanism that rewards moving first.

Interruption is not one behaviour

Speech that overlaps another person’s covers several quite different things. There is the supportive overlap, which agrees and encourages and is heard as engagement rather than interruption. There is the completion, where a listener finishes a speaker’s sentence. And there is the floor-taking interruption, which is an attempt to end somebody else’s turn early.

Conventions about which of these are acceptable vary a great deal between groups and settings, and the mismatch is a real source of friction. A person from a high-overlap conversational culture participating in a low-overlap one appears to interrupt constantly; the same person in reverse appears disengaged. Both are following a convention properly, in a room using a different one.

Remote meetings break the signalling, not the manners

Almost everything described here depends on precise timing and on peripheral vision, and both are damaged by a video connection. A delay of a few hundred milliseconds is enough to destroy the handover, because the cue that would have prepared the next speaker arrives after they needed it.

The result is the familiar pattern of two people starting simultaneously, both stopping, and both restarting. It is not a failure of politeness — if anything it is caused by an excess of it — and it is why explicit protocols work so much better online than in person. Raising a hand, using a queue in the chat, or having a chair call names are all crude compared with in-person turn-taking, and all of them outperform it when the signalling channel is broken.

Working with the mechanism

For anybody trying to contribute, the practical points are that the moment to enter is the approach to the end of a turn rather than the silence afterwards, and that a short signal of intent — a small gesture, an audible in-breath, leaning forward — is usually enough to be allocated the next turn without competing for it.

For anybody running a discussion, direct allocation is the single most effective tool available, and it is worth using deliberately rather than relying on an open invitation. It changes who contributes far more than any request for people to speak up, because it replaces a race with an allocation, and the outcome of a race in a group of adults is almost entirely determined by who is comfortable racing.