This study examined the direct and indirect effects between training in the Responsive Classroom® (RC) approach, teachers' uptake of RC practices, and teacher-student interaction quality, using a structural equation modeling framework. A total of 24 schools were randomly assigned to experimental or control conditions. Third- and fourth-grade teachers in treatment schools (n=132) received training in the RC approach, whereas teachers in control schools (n=107) continued "business as usual." Observers rated teachers' fidelity of implementation (FOI) of RC practices 5 times throughout the year using the Classroom Practices Observation Measure. In addition, teachers completed self-report measures of FOI, the Classroom Practices Teacher Survey and Classroom Practices Frequency Survey, at the end of the school year. Teacher-student interactions were rated during classroom observations using the Classroom Assessment Scoring System. Controlling for teachers' grade level and teacher-student interaction quality at pretest, RC training was expected to predict posttest teacher-student interaction quality directly and indirectly through FOI. Results supported only a significant indirect effect, β=0.85, p=.002. Specifically, RC teachers had higher levels of FOI of RC practices, β=1.62, p<.001, R2=.69. In turn, FOI related to greater improvement in teacher-student interaction quality, β=0.52, p=.001, R2=.32. Discussion highlights factors contributing to variability in FOI and school administrators roles in supporting FOI.