This paper describes student-level findings of the first large-scale comprehensive school effectiveness study of the primary education in Serbia. Twenty-five student-level variables were examined in a three-level HLM model using a study sample of almost 5000 students, over 250 classrooms and over 100 schools. Differences between the students were in large part responsible for differences in achievement scores in mathematics and Serbian language. Parental education, Roma minority status, developmental or family problems, gender, student motivation, parental involvement in student work and homework were some of the factors associated with student achievement. Serbian policy-makers are alerted to possible actions in order to improve mathematics and Serbian language achievement. (Contains 2 notes and 5 tables.)