As part of an ongoing investigation of the relationship between teacher behavior and pupil affective and cognitive outcomes, coders practiced recording their observations of existing classroom videotapes in order to determine the amount of interobserver agreement. The videotapes were coded, using seven existing low- to medium-inference classroom observation systems. All coders coded three reliability tapes prior to coding 36 test tapes. The systems used to code the videotapes were the Spaulding Teacher Activity Rating Schedule; Teacher-Child Dyadic Interaction System (Brophy-Good); Observation Schedule and Record Form; Flanders System of Interaction Analysis; Florida Climate and Control Systems; CERLI Verbal-Behavior Classification System; and Classroom Communication Observational Categories. Data on encoding, using two high-inference measures, the Classroom Observation Scales and Teacher Evaluation Scale, were also collected. Tables showing interobserver agreement on the 36 videotapes and coding directions for the observation instruments are appended. (MH)