A key argument currently invoked to cast skepticism on certain South American sites, that suggest a first peopling of the New World by ca. 35,000 BP, is the perplexingly low visibility of the archaeological record until 12,000 BP. But, contrary to a popular misconception, great spatial and temporal discontinuities are common in the Old World Palaeolithic settlement record. In southern Africa, carefully controlled archaeological stratigraphics show that the now semiarid interior was unoccupied for 50,000 and more years at a time. Episodes of widespread settlement in marginal environments were relatively brief, limited to periods of substantially wetter climate, and closely linked with moist habitats. A risk-minimization model is proposed to explain these discontinuities. Plant and animal resources in the region, given a climate as dry or drier than today, were of low productivity and low reliability for unspecialized hunter-gatherers during the dry seasons for poor years. This would require large foraging territories and very wide spacing of proximal bands, so that the exchange of vital information on temporary or migratory resources was minimal. Finally, during extended droughts, fat-depleted animals provided an unsatisfactory source of food. These variables suggest that environments with low productivity and predictability were too risky for unspecialized hunter-gatherers with a pre-Upper Palaeolithic technology, such as those who would have been able to enter the New World 35,000 BP. Major spatial and temporal gaps in the New World settlement record should therefore be expected prior to the appearance of specialized Palaeoindian hunter-gatherers ca. 12,000 BP. Implications for geoarchaeological strategies are discussed.