The development of a warm and supportive relationship with their teacher is protective for preschool children, and particularly beneficial for children who display early disruptive behaviors. Banking Time is a dyadic, short-term intervention to improve the quality of teacher-child interactions between a teacher and a specific child, building a more positive teacher-child relationship. During Banking Time sessions, both teacher and child interact toward an activity chosen by the child. This study used variable- and person-centered approaches to examine Banking Time implementation in a sample of diverse preschool teachers (N = 120) and children (N = 319) randomly selected to participate in the intervention. Results indicated that the majority of teachers implemented expected practices (i.e., observe and narrate the child's actions, allow the child to lead) as intended. However, although teachers were instructed to limit teacher-directed practices (i.e., ask questions, give praise, and use commands), teachers' use of these practices was not as limited as ideally expected. Three different implementation patterns were evident: (1) High Fidelity, in which the teachers engaged in the child's activity while ensuring that the child led the session; (2) Low Engaged, in which the child led the session, but the teacher was not engaged in the child's activity; and (3) Teacher-Led, in which the teacher was engaged in the child's activity but also directing the session instead of ensuring the child's lead. These implementation patterns were linked to changes in the quality of teacher-child interactions. Implementation findings from this study can inform interventions to improve the quality of teacher-child interactions. (author abstract)