Did the Third International Mathematics and Science Study (TIMSS) ask eighth grade science teachers the right questions about their use of instructional time? TIMSS asked teachers to recall a lesson that they had taught, and then group activities into 11 categories. This study examined the TIMSS question "How did the lesson proceed?" by videotaping six classes of eighth grade science in Alabama and Virginia and comparing observer coding of the video to the teachers' recalled descriptions of the same class. The manner in which the TIMSS data were collected and the manner in which data were collected from teachers in this videotape study suggested the use of a repeated measures analysis of variance (ANOVA) model. Using a repeated measures analysis allowed us to look at the interactions between teachers and observer and the 11 TIMSS activities, the 26 NSF student activities, and the 11 NSF teacher activities. The difference between observer and teacher responses using TIMSS categories was not significant; however, 43% of the total variance was explained by whether the teacher or the observer reported the times for the instructional activities. The teachers also responded to questions from the NSF Local Systemic Change Through Teacher Enhancement K-8 Teacher Questionnaire to describe the same class. The difference found between the teacher and the observer coding was not significant, but the amount of variance explained by the data source (observer or teacher) dropped to 33% when using NSF student activity categories and to 26% when using NSF teacher activity categories. The study concluded that questionnaires to survey science teachers about their instructional activities should include operational definitions, methods of classifying single activities into multiple instructional categories, and questions that are more accurate in describing quality science instructional activities. (Contains 43 references.) (Author/SAH)