This paper is concerned with key concepts in conversation analytic research: projection, conditional relevance and the adjacency pair. Focusing on a specific type of adjacency pair (instructing/instructed action) that constitutes an exemplary case for the analysis of embodied practices, the paper examines how verbal and embodied resources are mobilized, temporally organized and oriented to by the participants to form instructing and instructed actions.
A recurring practice in my data is the combined use of the German deictic form "so" with concurrent bodily demonstrations. In instructional sequences, the expression "so" not only serves as a turn-internal flag (Streeck, 2002) that directs the addressee's visual attention to the speaker's bodily activities, but it can also be part of a projecting first (instruction) or of a projected second pair part (instructed action). Drawing on video recordings from different settings, it will be shown how demonstrations carried out by the instructing person on the one hand and by the instructed person on the other hand are embedded within the sequential format of instructions. The paper argues for a multimodal conceptualization of central notions of CA such as turn, adjacency pair, and conditional relevance. It also raises the theoretical question of how the concept of conditional relevance as it defines the classical adjacency pair relates to the concept of projection when both are (re)considered from a multimodal perspective.