Classroom observation is one of the premiere data collection methods available to those interested in teaching behavior. Observational techniques can be classified on a continuum ranging from low inference to high inference depending on the level of judgment required by the observer making the observation. Central to the issue of any form of measurement are score reliability and validity. This paper explores various observational methods and discusses related reliability and validity issues. These include observer biases, intrusiveness of the observer, observer training, and contextual issues. Validity concerns explored include those of face, content, construct, predictive, and observer validity. Some of the key studies that use classroom observation instruments are described. Further research is suggested based on the "training" factor relevant to the classroom observation. (Contains 31 references.) (Author/SLD)