Issues in collecting data on the classroom performance of teachers are discussed, with a focus on written descriptions of teacher behavior in contrast with videotaped observations. The central issue in this comparison is the comparability of the information when it is observed live and when it is simply recorded on videotape. The fundamental psychometric questions associated with any assessment technique can also be posed in terms of validity, reliability, and bias. In any data collection effort, the purposes must be clear, and the inferences to be drawn must be recognized. The situational generalization of observations is limited in different ways for live observers or videodocumentation. Live observers restrict their attention, and videorecorders may restrict the observers' focus. However, the videodocumentation of classroom teaching produces a record that can be viewed, interpreted, and evaluated by many observers. Limited research from J. R. Fredericksen and colleagues suggests that live observation and videotape scoring produce very similar evaluations of teachers, with adequate interscorer reliability. Various practical issues in videotape use, and opportunities it provides for professional development are discussed. Research needs are reviewed. (SLD)