A multilevel model of achievement in science using data from the OECD 2006 Programme for International Student Assessment (PISA) is presented. School-and student-level variables are analysed as conceptually-related blocks. A combined model explains almost half of the total variance in achievement (42% within schools and 79% between schools). Over and above student-level variables, school-level variables explain just 1% of the variance. Of the total explained variance, about 5% is accounted for by students' self-reported efficacy in science and enjoyment of science. The model confirms contributions of student and school socioeconomic status, home educational climate, take-up of science for the Junior Certificate, and general engagement in science. For students not taking science, boys have lower expected scores than girls.