A generalizability study of timed observations was conducted in 25 primary grade classes to observe teachers' use of time--for instruction, evaluation of instruction, and classroom management--according to the hour and day observed. Observational methods used by on-site researchers included videotape, checklists, running documentaries, frequency counts, and continuous time allocation by category of behavior. Critical elements of classroom observation were determined to be the teacher, the situation, the occasion, and the evaluator. Teacher behavior in each of ten categories was observed: (1) clerical; (2) preparation of facilities; (3) preparation of materials; (4) planning for instruction; (5) instruction; (6) evaluating instruction; (7) classroom management; (8) supervision; (9) administration; and (10) not teaching. This study appeared to confirm results reported elsewhere: that a total of between four and six observations of teacher behavior, separated over several days, should provide relatively reliable measures of teacher behavior. (MH)