Validity is a central issue in use of self-report data for the investigation of teacher thinking. "Triangulation," or use of several data sets to test, corroborate and elaborate each other, is one method for increasing validity of findings. This paper illustrates a process of triangulation of data obtained by three different data collection procedures: stimulated recall interviews; Kelly Repertory Grid interviews; and ethnographic observations of classroom interaction. Comparative case studies of two junior high school teachers are presented, and data are analyzed to provide information on teachers' cognitive processes and beliefs. Two new procedures for analyzing stimulated recall data are introduced. Starting with the stimulated recall data, each data set is examined in turn, and initial findings are tested, extended, and revised with the addition of each new data set. The potential power of the triangulation method is demonstrated, and potential problems are identified. (Author)