Fertility in South Africa has been falling for almost four decades. The 2001 South Africa Census offers the opportunity to reflect on this decline, and to assess the trajectory and patterns of fertility in the country, among its population groups and in its provinces. Analysis of the data in the 2001 census shows that fertility among all four main population groups continues to fall, and that the national level of fertility is now below three children per woman. The rate of decline indicated by the estimated levels of fertility is a continuation of the long trend of gradually declining fertility. The 2001 census was the second conducted in a postapartheid South Africa. The first, which was conducted in 1996, is regarded as the most reliable and accurate enumeration of the South African population since that in 1970. The comparison of fertility levels and trends estimated from these two post-apartheid censuses provides valuable checks and comparisons that further enhance our understanding of fertility dynamics in the country.Nationally, fertility has fallen by 0.4 of a child per woman over a five year period, a decline of 12.1 per cent. The rate of decline has been fastest among Indian/Asian women, and slowest among Coloured women. The level of, and the rate of change in, provincial fertility reflects both the composition of each province by population group as well as differences in the level of urbanisation. Fertility is lowest in the Western Cape, Northern Cape, Gauteng and the Free State. These four provinces also showed the lowest fertility levels in the 1996 census. A particularly large decline in fertility is apparent in the data for the Eastern Cape. Whether this fall is ‘real’ or – more probably – a result of errors in the data collected in that province isunknown at this stage. Further research, based on household survey data, together with the Demographic and Health Survey (DHS) that is currently in the field, may be able to shed light on this question. The derivation of robust and reliable fertility rates is a worthwhile endeavour in its own right and represents the major portion of the present work, but to do so requires that this report\ndiscusses the quality of the data collected in the 2001 census at some length. We identify (and correct for) several significant anomalies in the data collected in the 2001 census. The two errors that give rise to the greatest concern are, first, the apparent inability of the\ncensus to capture accurately all births that occurred in the preceding twelve months. As with the 1996 census data, only about half of the 1.1 million births we estimate to have occurred in the country over the year before the 2001 census were actually enumerated. Errors of this sort are common, and well documented, in all developing country settings. Nonetheless, strenuous efforts should be made to improve the quality of recent fertility data collected in future censuses. Second, the data on the number of children born to women of reproductive age are seriously deficient. All indications are that this was a result of inadequate (or incorrect) training of enumerators. This matter can, and must, be addressed before the next census if these data are to be useful. While a method is presented here that allows estimates of women’s lifetime fertility to be calculated, the ramifications of this flaw in the data extend beyond the analysis and determination of fertility levels and trends. Most importantly, the data on the proportion of children born that are still alive indicates that (most probably) women interpreted the question on children ever born as asking about children still alive. This compromises the estimation of child mortality levels from the census data, since the estimation procedures rely heavily on those proportions. Further investigation of these data is outside the scope of this report, but is covered in a second monograph prepared by the Centre for Actuarial Research for Statistics South Africa (Dorrington, Moultrie and Timæus 2004).