Tickle-Degnen and Rosenthal identified the nonverbal correlates of rapport — mutual attentiveness, positivity, and coordination — and the codes here are the frame-observable members of that set. Unlike the other image scales, these are dyadic: they are properties of a pair, not of one person, which is what makes them the right unit for supervision review.
Intended for detection studies. Draw a box that contains both participants (or both relevant regions) — the code is about the relationship between them, so a box around one person cannot support it. That is a deliberate departure from the per-person scales and the most common coding error with this scheme.
Why these four. Each is a coordination signal rather than an individual display: gaze that is returned, a nod that answers something, a posture that echoes the partner's, a smile that is reciprocated. An unreturned smile is not an engagement cue — it is an affect display, and belongs to Facial Affect (Image Coding).
Mirroring needs care. Two people in similar postures because they are sitting in identical chairs is not mirroring. Code it only where the correspondence is in a configuration the setting did not impose.
Severity is clarity. Reciprocity is a two-sided judgment, so if either participant is occluded, cropped or out of focus, the ceiling is Partial. This is the scale where a wide, low-resolution room shot is most limiting.
Nods and mirroring are movements. A single frame gives weak evidence for both; prefer video where the distinction matters, and rate honestly when it does not.