We examined teacher supervision practices (supervision models, phases, and professional development guidelines) of in-school supervisors (principals, vice principals, and study program directors) in Uganda, the supervisors' efficacy perceptions regarding teacher supervision, and supervisor characteristics associated with the choice of supervisory practices. A quantitative survey on instructional supervision was conducted among in-school supervisors in 45 secondary schools in three rural school districts in Uganda. This country faces the challenge of assessing the quality of its education, in light of reforms in universal primary and secondary education programs, and of improving student achievement through teacher professional development. A total of 106 supervisors responded to the questionnaire. Factor analysis, descriptive statistics, correlations, t test analysis, and analysis of variance were performed. Results indicate that the supervisory practices most used were a differentiated model and a 360° supervision model, a supervision phase prior to class observation, and a phase following class observation, as well as directive and collaborative professional development guidelines. The supervisors' perception of collective efficacy was stronger than was their perception of self- or professional efficacy, and their internal perception of efficacy was generally stronger than was their external perception of efficacy. Supervisors' sociocognitive characteristics, particularly the internal perception of professional efficacy, were the factors most associated with the choice of supervisory practice. Our findings suggest that the choice of supervisory practice is greatly associated with the supervisors' perceptions of efficacy. Before introducing change, we must learn more about the practices and perceptions of the teaching supervisors. This study addresses this aspect.