Reported is a study on teacher's visually perceived nonverbal behaviors and their relationship to student's affective responses among 30 junior high teachers. Each teacher was requested to present a lesson at the front of the classroom. About 20 minutes of one of their classes were videotaped. Two 5-minute segments (episodes) and one 15-second segment of teacher behaviors were edited and extracted. Coding symbols were used to identify and categorize individual behaviors by different observers, and interobserver agreement coefficients were computed by Scott's method. For each behavior, data on frequency and duration of occurrence were collected. A total of 879 students in a remote community was asked to react to five episodes, and 395 students, in addition, reacted to brief tapes. A semantic different scale was used to record responses, and teacher characteristics were described by students. Response means were computed for boys, girls, and boys and girls combined. Computer analyses showed that nonverbal behaviors measured in terms of duration are good predictors of student affective responses. Students preferred some kinds of teacher behaviors. Boys and girls responded differently to teachers. Students tended to make their affective judgments in the absence of consciously perceived awareness. Included in the appendix are categories of teacher nonverbal behaviors. (CC)