Three cave sites, Nelson Bay Cave on the coast, Kangkara in an intermontane valley and Boomplaas some 80 km inland in the southern Cape Province, South Africa, were excavated between 1970 and 1979. Nearly 225 000 stone artefacts from the Later Stone Age sequence dating within the last 20 000 years are described from these three sites and are related to micro- and macroevolutionary changes at a regional and sub-continental level to provide a model for change in the Later Stone Age. The classification scheme was designed to highlight inter- and intra-site variability through time and focused on analysis of the successive stages in the reduction sequence from raw material nodule to finished tool. Linear regression was used to test for interdependence and independence of variables and the significance of changes in size and shape of untrimmed flakes and scrapers was assessed. In the case of the scrapers, samples from the three southern Cape sites and from the eastern and northern Cape were compared and tested for significance with Mann Whitney and Kolmogorov-Smirnov two-sample non-parametric tests. The results indicate two levels of change through time, that involving the appearance of innovations that can be described as macroevolutionary, and that involving the subsequent modification of the frequency, size and shape of these innovations and other artefacts already part of the toolkit that can be described as microevolutionary change. By comparison with modern technological data, innovative changes represent the diffusion and acceptance of major advances in tool technology that are predictable from trends observed in modern technology. Post-innovative or microevolutionary changes, on the other hand, take the form of oscillations around a gradually changing mean that are similar to changes in style or fashion in the modern idiom. The hypothesis that technological change was stimulated by environmental change was tested by comparing the timing of technological changes with those seen in oxygen isotope ratios in a Cango Cave stalagmite, charcoals from woody plants brought into Boomplaas Cave for firewood, small mammals caught by owls and eaten at Boomplaas and Nelson Bay Cave, and larger mammals hunted by people at all three sites. The results indicate that there is no consistent relationship between changes in the stone tool technology and environmental change. There is, however, a coincidence in the timing of changes in the larger mammals hunted and the stone tool technology that took place over a relatively short span of time between 12 000 and 11 000 B.P., post-dating major environmental adjustments at the end of the last glacial cycle by some 3000 years. Technological changes that took place between 8000 and 6000 B.P. were not coincident with a change in the animals hunted, nor with an equally sudden shift in environmental data, while a change in economy from hunting to herding within the last 2000 years was not accompanied by a change in the stone tool technology although pottery was added to the toolkit. There is thus a very complex relationship between economy, technology and environmental change that is not readily predictable. The sequence in the southern Cape can be described in terms of punctuated equilibria, but the times of rapid change in technology, economy and climate do not always coincide. In the technological system periods of relative stasis have been labelled the Robberg, Albany and Wilton industries. The content, dating and evidence for subsistence during the Later Stone Age south of the Zambesi is reviewed from several hundred dated horizons at over 160 sites and although there is some spatial variability, the sequence of technological changes is much the same throughout the sub-continent. This confirms the long-held belief that the innovations that spread through the sub-continent were diffused over a very wide area of the Old World as the result of a well developed network of intercommunications during the Stone Age, while at times population migrations also took place. Microevolutionary changes, on the other hand, tend to be more regionally specific and may have been stimulated by different cues.