The rationale for and development of the Classroom Behavior Scales (CRB) are discussed in connection with applications on four separate samples of Head Start, kindergarten and first grade children. Studies of observer agreement, stability, structure, comparisons over sex, classes, conditions, observers and over time are presented. Results showed high observer agreement, non-significant repeated trail correlations, consistent and expected changes in scale means over time--Head Start to first grade, unstable factor structures over samples and over repeated observation of the same sample. Differences between classes were consistently significant. An application with an experimental program did not show significant differences, which was an expected result. Classroom observational strategies are compared to educational and psychological testing as ways of getting measures of children in schools. While the latter will, in general, put a premium on stability, as a condition of test reliability, the former will emphasize individual variability. Data from CRB applications is presented to support this contention. The key for the scales of the CRB is appended. (Author)