The dynamics of the give and take between therapist and client(s) is frequently of interest to therapy process researchers. Characterizing the ways that therapists respond to clients and clients respond to therapists can be challenging in therapeutic encounters involving a single therapist and a single client. The complexity of this challenge increases as the number of people involved in a therapeutic encounter increases not only because there are more people responding to one another but also because the patterns of responding can become more complex. This manuscript demonstrates how dyadic cross-lagged panel models can be extended to psychotherapeutic encounters involving 3 people and used to test processes that exist between dyadic subsets of the larger group as well as the group as one cohesive unit. Three hundred seventy-nine talk turns of fundamental frequency from a couple therapy session were modeled using 3 dyadic cross-lagged panel models, and each individual’s respiratory sinus arrhythmia (RSA) was treated as a moderator. Although the regression coefficients for each dyadic subset (e.g., therapist–husband) were nonsignificant, an eigenvalue/eigenvector decomposition of the regression coefficients from the 3 dyadic cross-lagged panel models suggests that interdependence exists at the level of the whole group (i.e., therapist–husband–wife) rather than between pairs of individuals within the group (e.g., husband–wife). Further, an interaction involving husband’s RSA suggested that interdependence involving the husband ceased when the husband displayed greater regulatory effort. This combination of statistical methods allows for clearly distinguishing between dyadic therapeutic processes and group-level therapeutic processes. (PsycINFO Database Record (c) 2017 APA, all rights reserved)