A striking feature of labour supply in South Africa is the phenomenal expansion in the labour force participation of women from 38 percent in 1995 to 46 percent in 2004. Even so, their participation has been persistently lower than that of men whose participation rates were 58 percent and 62 percent respectively. Furthermore, analyses of women's participation rates by race show that the rates for historically disadvantaged groups such as Africans are still lower than those of Whites. For instance, in 1995 African women had a participation rate of 34 percent and it increased to 43 percent in 2004 while the corresponding rates for White women were 52 percent and 59 percent. In light of these disparities, this paper uses survey data to examine the determinants of the low level and also of the changes in African women's labour force participation, during the first decade of democracy (1995-2004). By focussing on a ten year period, this research substantially differs from earlier studies which were preoccupied with short periods such as one year. A longer period is analytically advantageous because it allows the capturing of the changes and the robustness of the key determinants of female labour force participation in South Africa. Such information is important not only for reviewing existing policies but also for the formulation of new ones to increase female labour force participation which is a prerequisite for economic development. The study utilises a decomposition technique devised by Even and Macpherson (1990). The findings exhibit that female participation responded positively to education which has been the prime factor. Non-labour income, marriage, fertility and geographical variations in economic development persistently stifled participation. It is argued that the perceived change in participation is due to emigration and changes in human capital and financial endowments. Another important discovery is that -9 percent of the observed shifts in the participation rates from 1995-2004 is due to disparities in characteristics while differences in coefficients account for 109 percent of the shifts.