This study examined the relationship between observed classroom behavior (teacher-pupil interactions) and reading achievement. An elementary school reading teacher and six students with different coping styles were observed six times in each of 41 classrooms during the school year. Pretest and posttest reading scores, a measure for socioeconomic status, grade level, and scores from 21 categories of the "Spaulding Teacher Activity Rating Schedule" (STARS) were the variables used. Results indicate that the pretest is the best indicator of reading achievement (accounting for almost two-thirds of the variance); that several of the STARS scores are better predictors of reading achievement than either grade level or socioeconomic status; that a negative relationship exists between reading and some teacher/pupil interactions that were supposed to be "good," such as structuring supervisor and supportive guide; and that a positive relationship exists between reading achievement and egocentric instructor. These findings suggest that, although the pretest is the best indicator of a student's growth in reading, observed behaviors of both students and teachers may be useful predictors of reading achievement. (RL)