This paper aimed to highlight the process of therapist direct contingent responding to shape client behavior in two Child Behavior Analytic Therapy (CBAT) cases using transitional probabilities. The Functional Analytic Psychotherapy Rating Scale (FAPRS) was used to code client behaviors and the Multidimensional System for Coding Behaviors in Therapist-Client Interaction (SiMCCIT) was used to code therapist behavior. Measures of behavior duration and transitional probability were taken from nine sample sessions for Client 1 and 10 sample sessions for Client 2. The first treatment was successful and lasted 18 sessions, while the second treatment was unsuccessful. In the first case, coding results showed that problem behaviors (CRB1) decreased, and improvement behaviors (CRB2) increased. In the second case, CRB1 predominated through seven sessions, and CRB2 did not show stability. Transitional probabilities indicated regularities in therapist responding to the client's clinically relevant behaviors, especially with approval after CRB2. It is concluded that contingent responding with approval functioned to shape client improvements in the first case, but not in the second case. (Contains 2 tables, 2 figures, and 1 footnote.)