Skip to content

Turn construction unit

An increment of talk that participants can project as possibly completing a speaker's turn, thereby creating a transition-relevance place where speaker change may occur.

Version
v1 · 2026-09-28 · History
Domain-specific #
12657
Domain group
Social Sciences
Origin domain
Sociology & Anthropology
Subdomain
Conversation Analysis → Sociology & Anthropology

Core Idea

A turn construction unit (TCU) is an increment of talk that participants can project as possibly completing a speaker’s turn. Its first possible completion creates a transition-relevance place, where the current turn-taking system permits selection of a next speaker, self-selection, or continuation by the current speaker. Speaker change is merely possible there, not automatically required.

TCUs can be built through lexical, phrasal, clausal, sentential, and other interactionally usable constructions. Their identity is not purely grammatical. Participants draw on grammar, prosody, action, gaze, and sequence context to project what the unfolding contribution may require before reaching possible completion.

Scope of Application

TCUs are used in conversation analysis to examine ordinary talk, institutional interaction, repair, turn allocation, overlap, and response timing. A turn may contain one TCU or continue through several.

The concept is not identical to a sentence, pause, transcript line, or entire turn. It can be used across languages and signed interaction only when analysts establish the locally relevant resources for projectability rather than impose English sentence boundaries.

Clarity

The abstraction separates constructing talk from allocating speakership. It explains how recipients can prepare to enter before a contribution ends, why transitions cluster around recognizable completion points, and why a current speaker may continue after reaching one. A transition-relevance place is an opportunity produced by possible completion, not proof that transfer occurred.

That distinction matters when annotating recordings. A silence after a grammatical clause may not mark completion if intonation and action project continuation, while a short lexical item can form a complete TCU in the right sequence. Analysis must demonstrate how the parties treated the unfolding talk rather than declare boundaries from written syntax alone.

Manages Complexity

Moment-by-moment interaction combines syntax, intonation, embodiment, social action, and recipient response. TCU analysis organizes these details around a focused question: what can participants now project as possibly complete? The unit compresses evidence for timing without reducing interaction to grammar alone.

Abstract Reasoning

Identify the developing action and candidate construction. Examine what participants could project before the apparent endpoint, using grammar, prosody, embodiment, and sequence. Locate the earliest defensible possible completion and inspect recipient conduct: entry, withholding, overlap, or continued attention. Separate the analyst’s transcription choice from participant orientation, and allow extensions, repairs, and additional TCUs to revise the trajectory.

Knowledge Transfer

The framework transfers across interactions when the receiving setting supplies its own completion resources. The English inventory of unit types does not travel unchanged to every language or modality. Outside conversation analysis, an analogy is sound only when agents project completion and coordinate allocation around it. Mere segmentation into convenient units does not instantiate the participant-oriented abstraction. The receiving analysis must preserve possibility, projectability, and locally demonstrated turn-taking consequences together.

Neighborhood in Abstraction Space

Turn construction unit sits in a sparse region of the domain-specific corpus (93rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08