This paper discusses findings from four years of observational research examining classroom climate. Specifically, findings that focused on one powerful teacher behavior (teacher retreating) were examined. These data indicated that teacher retreating was negatively correlated to the amount of instructional interaction occurring between students and teachers and was positively correlated to the amount of unwanted student behavior observed. Regression analysis indicated that teacher retreating was the only behavior that predicted the amount of student off-task behavior taking place. The authors discuss the importance of these findings to teacher educators and offer suggestions for preventing teacher retreating.