Individual pupils were observed in junior high school science classes to characterize and record their classroom behaviors. The sample (N=80), drawn from five urban schools, was evenly divided by race (black and white). Instruction in class was based on laboratory work or other activities requiring active pupil participation. Systematic observations made by trained observers over a 3-month period yielded frequencies of 14 behaviors that were combined to form 3 broad behavior categories: (1) active learning; (2) passive learning; and (3) non-attending. No systematic race or sex differences were found in behavior. Entering math and reading scores were the best predictors of final grade but active learning behavior was also a significant contributor to final grade. Pupils tended to interact most frequently with members of the same race. The most significant finding is that, for the junior high school pupils in this sample, obstacles to achievement in science were not related to race and sex, but rather to poor math and reading skills and failure to participate in the learning activities of the classroom. (Author/JN)