A study explored the use of an interactive analysis system, the Reciprocal Category System (RCS), for recording and analyzing observations of individuals tutoring in reading. The supervision of teaching requires observation, and the RCS can provide detailed, objective, and quantifiable records of behavior categories that occur in teacher-learner interactions and indicate whether the behaviors subsequently diminish or increase. Three observers first viewed videotapes of tutoring sessions. The data were then used to determine intraobserver and interobserver reliability. Five aspects of instruction were selected to see if they could be predicted from interaction patterns. It was found that the RCS categories are not sufficiently specific, in that types of questions are not identified, and correct and incorrect responses are not discriminated. Due to the small number of comparisons, the usefulness of certain indices is tentative. Refinements in RCS categories are called for before further research is undertaken. (CRH)