The researchers concentrated on developing appropriate statistical models and procedures for analyzing three basic types of data used to describe teacher behavior: behavior count data that records the frequency of a given behavior; Bernoulli trial data in which the number of times a behavior occurs is counted against the total number of opportunities for its occurrence; and quantitative measures, including high- inference measures, derived quantities, or a quantity. The statistical procedures to measure the consistency over time of the behavior of an individual teacher must assess the viability of the relevant homogeneity hypothesis and estimate the amount of heterogeneity. Three approaches to assessing stability among teachers have dominated previous research: computation of correlations among observation times, application of generalizability theory, and estimation of occasion effects in repeated measures analysis of variance. Finally, the authors formulate two related research questions for stability across different contexts, i.e., subject matter or class composition. 10 tables, 3 figures, 11 notes, 82 references, and 2 appendixes