During the 60's, systems for the observation of classroom behavior have become an important area of educational research. Many new systems have been devised both for the evaluation of teaching techniques and for basic research into teaching processes. In the development of such systems the investigator must determine what behaviors are to be observed. This in turn requires such considerations as dimensions of classroom behavior, types of observation systems, and units of behavior. The second task of the investigator is to decide how the observations are to be carried out, e. g., recording and coding of behavior, reliability of coding, and statistical analysis. Three illustrations of observation systems are: Flander's system of classroom interaction analysis (affective dimensions), Bellack's observation of classroom discourse system (cognitive dimensions), and Oliver and Shaver's observational system for analyzing styles of teaching (affective and cognitive dimensions). The range of applicability of many observation techniques is limited and many investigators feel the need to devise their own. This has its drawbacks for comparisons even though the current state of the art often makes it necessary. (DJB)