This article aims to investigate variations in science achievement for secondary school students across the Trends in International Mathematics and Science Study (TIMSS) assessments and to examine the relations of several student- and school-level factors with achievement. The data were obtained from 5,577, 5,314, and 4,466 Malaysian eighth graders who participated in TIMSS 1999, 2003, and 2007, respectively. A two-level multilevel modeling technique was employed. The results indicate that 48.64%, 48.96%, and 60.95% of variance in achievement scores accounted for between-school differences in the 1999, 2003, and 2007 studies, correspondingly. After controlling for the school-level factors, gender followed by the science self-concept in both the 1999 and 2003 studies; time spent on working at home and valuing science in the 2007 study, yielded the strongest link to achievement. At the same time, controlling for the student-level factors, school location, instructional limitations, and teacher emphasis on homework had the strongest association with achievement, respectively, in all three studies. It was concluded that the variation in science achievement at both student and school levels was greatly increased in TIMSS 2007 compared to the two previous studies. (Contains 1 figure and 4 tables.)