Written to assist future researchers in conducting cross-national studies of classroom behaviors, this study indicates a number of factors which may reduce the validity of comparisons made. It is based on an investigation which explored the feasibility of using the Activities Categories Instrument and the Flanders Interaction Analysis Categories as a basis for making valid comparisons between classroom behaviors in the United States and Nepal. These instruments consist of categories of behavior, recorded by an observer, which can be used to describe teacher and student talk. Some of the 15 factors discussed include: teachers and students may behave differently than normal, given the presence of an observer; since category systems make use of time units to standardize the frequency or pacing of an observer's recording, a certain amount of distortion takes place because classroom events may break on or between the arbitrary time boundaries; researchers may encounter problems associated with similar but functionally different behaviors, and functionally equivalent but different behaviors; and the failure of personnel to follow a project's planned methodology either unconsciously or as a result of dishonesty. It is concluded that further research is needed to explore the strengths and weaknesses associated with category system usage. (Author/JK)