Objectives: This study of a field instructor (FI) training model, offered at two universities, focused on the relationship between student competencies, the supervisory alliance, and students' attachment styles. Method: The study used a pretest-posttest follow-up design of 100 randomly assigned FIs (training group = 48; control group = 52) and 64 students. The authors used Hierarchical linear modeling (HLM) to estimate linear growth models and the relation between variables at different times within and across supervisory dyads. Results: The training improved the supervisory relationship and selected student competencies as rated by FIs but not by students. These findings were not moderated by student attachment style; however; relationships were found between students' anxious and avoidant attachment styles and student-rated performance. Conclusion: This randomized controlled study underscores the need for future research in developing and testing FI training models. (Contains 6 tables.)