There are thousands of prehistoric rock art sites throughout southern Africa. Most of the paintings are attributed to Stone Age hunter-gatherers, the ancestors of the San (‘Bushmen'). For many years these images have been the subject of intense scrutiny, with the focus on the nature of the subject matter, the meanings of the art, and possible dating techniques. This study was conducted to investigate a previously unresearched aspect namely the chemistry and mineralogy of San paints in order to determine their composition. The obvious benefit resulting from this research would be the development of improved methods of rock art conservation. Additional benefits might include the establishment of a system that would aid in relative dating, the identification of differing painting technologies related to both time and space and, possibly, the identification of trading routes. Maqonqo Shelter, located approximately thirty-five km south east of Dundee, KwaZulu-Natal, was chosen as the primary study site because it contains a large number of paintings, though many are poorly preserved. Thirty paint and three blank (unpainted shelter wall) samples were analysed. Three additional sites were also sampled for comparative purposes, namely Twagwa Shelter (south coast), Sheltered Vale (southern Drakensberg) and Fergies Cave (central Drakensberg). Where possible, three blank, and three red and white samples were collected at each of these sites. All the samples were mounted on aluminium stubs, and analyzed in a Phillips XL30 environmental scanning electron microscope (ESEM), fitted with an EDAX detector. The samples were also analyzed using synchrotron radiation at the National Synchrotron Light Source at Brookhaven Laboratory, Long Island, New York. Both the mineralogical and minor elemental compositions of the samples were determined simultaneously using this technique, enabling over 400 analyses to be collected. A comparison of these datasets highlighted the differences in composition of the colours both within and between the individual sites. It has been shown that the intensive analysis of micro-samples of even ‘poor quality' paint samples, using a number of complimentary techniques, allows for the sourcing of the samples to individual areas. The determination of the inorganic component of the various paint colours has also indicated possible sources of the various pigments utilized e.g., orange is probably sourced from bauxite, red from haematite (not heated goethite), and white from whewellite (not kaolinite). Variations in the minor elements in the various colours (both within and between the different sites) have also indicated a wide variety of sources for these pigments, and not one source as has been previously postulated. The mineralogical nature of the paints gives an important insight into the sensitivity of the individual colours to weathering, and indicates that the conservation practices currently employed should be modified to be site, and sometimes colour, specific. It must be stressed that these findings are specific to the four sites studied, and are not necessarily representative of all San rock art. This preliminary study has shown the applicability of the methodologies used, and the findings will serve as a basis for future research in this field. Such further work could involve the analysis of paint samples in relation to the style of the paintings, as well as the determination of any organic components present.