Description: South Africa's education system is still deep in the throes of reform under its third Minister of Education since 1994. However, it is marked by underachievement of pupils at each level of the system. Poor communities, in particular, those of rural Africans bear the brunt of the past inequalities and these are continuously reflected in the national results of the final year examinations in Grade 12. The South African education system accommodates in the order of more than 12.3 million pupils (50.5% female). Equity and access are at the top of the government's priority list and access has improved to the extent that primary education is almost universal However, only 86% of South African pupils are enrolled at secondary school, despite the fact that education in South Africa is compulsory and supposed to be free for grades 1 to 9. Pupils are only expected to pay fees for grades 10 to 12, but educational user fees are widespread across all the grades. South Africa participated in TIMSS 1995, 1999 and 2003. Secondary analyses of these studies is underway and has revealed the large inequities in the education system with 50% of the variance in the pupils' mathematics scores explained between schools (Howie, 2002). However, this is mostly explained by the historical inequities imposed on communities and schools over the past forty years. The challenge is to explore the extent of the "gap" in pupils' scores by comparing the advantaged and disadvantaged communities in this context, namely the former in well-resourced, largely urban schools and the latter, largely under-resourced, black rural schools. Previous work conducted for mathematics pointed to crucial student-level factors such as language, Socio-economic status, perceptions of maths, attitudes towards maths, and self-concept (see Howie, 2002). This paper aims to ascertain the extent to which these factors have an effect on science and to what extent other factors play a role. The TIMSS 1999 data is explored to focus on the extent of the gap in mathematics achievement between advantaged and disadvantaged communities and specifically the factors that predict the outcomes in science in both communities. Linear equation modelling is applied to explore the latter to isolate these factors. Whilst the secondary analyses conducted for science so far on the South Africa data (Howie and Scherman, 2005, Howie, Scherman and Venter, 2004) revealed that many of the factors on schools and classroom level were the same, the extent to which they had an effect on the science score differed greatly in some cases. It is believed that on student-level this will certainly be the case and in particular between the two communities described above.