This study examined the effects of motivation to learn science, science self-beliefs, and science instructional practices on science achievement of 13,985 15-year-old students from 431 schools across Canada. Hierarchical linear modeling (HLM) analyses, while controlling for student- and school-level demographic characteristics, revealed the substantial predictive effects of motivation to learn science, science self-beliefs, and science instructional practices on science achievement of adolescents. Motivational beliefs--self-efficacy and self-concept--and enjoyment of science had substantial positive predictive effects on science achievement. In contrast, general interest in science had a negative predictive effect on science achievement in the context of other variables. Whereas science teaching using hands-on activities had a substantial positive predictive effect on science achievement, science teaching using student investigations had a substantial negative predictive effect in the context of other variables. The final HLM model indicated that only 8% of the variance in science achievement was between schools and 92% of the variance involved students within schools.