Objective: In a study of cognitive behavioral therapy (CBT) for generalized anxiety disorder (GAD), greater early patient-therapist convergence on post-session perceptions of their shared alliance quality was associated with better subsequent outcomes (Coyne et al., 2018). To further understand this evidence-based process, the present study examined whether in-session interpersonal microprocesses differentiated dyads known to possess high versus low early alliance convergence. First, I hypothesized that high versus low convergence dyads would be characterized by more overall positive interpersonal complementarity; moreover, in light of an interpersonal vulnerability associated with GAD (i.e., high deference to others), I investigated whether a specific positive complementary sequence (i.e., the patient asserting and separating from the therapist’s influence, and the therapist responding with affirmation and understanding) was more representative of high versus low convergence. Second, given their proclivity for misunderstanding and frustration (e.g., mixed messages of friendliness and hostility), I hypothesized that fewer complex communications would characterize the high versus low convergence dyad. Finally, given its generally positive influence on close relationships, I predicted that therapists in the high versus low convergence group would engage in more self-disclosure. Method: I selected 8 high and 8 low convergence dyads from Coyne et al.’s (2018) sample. The index of early alliance convergence was derived from patient and therapists’ session-by-session ratings (from sessions 2-8) on their respective version of the Working Alliance Inventory-Short Form. To assess the interpersonal processes across 3 of these early-treatment sessions, independent raters applied the Structural Analysis of Social Behavior. Results: Counter to my hypothesis, there was no difference in positive complementarity between the alliance convergence dyads. However, as predicted, the mean probability of the specific positive complementarity sequence occurring in the high convergence dyads was 0.79 SDs greater than in the low convergence dyads. Also counter to my hypotheses, there was no difference in the frequency of complex communications or therapist self-disclosures between the alliance convergence dyads. Conclusions: For people receiving CBT for GAD, alliance convergence patterns—a dyadic factor that explains unique variance in patient improvement—may be distinguished more by specific, pathology-relevant patient-therapist sequential exchanges than overall levels of interpersonal processes