The age and properties of detrital zircon in a sediment reflect properties of the rocks in which the zircon crystallized, and not necessarily the immediate precursor of the host sediment. Where clastic sediments are recycled, these properties are preserved so that U-Pb ages and Lu-Hf data on zircon grains will no longer give information on the routing of detritus. Because the crustal evolution and sedimentary recycling history of Southern Africa is well-established, Holocene sediments in the region provide a good test example for detrital zircon geochronology applied to sands or sandstones with a complex recycling history. Data from 16 samples of unconsolidated sand, including dunes from the southern part of the Kalahari Basin, beach sands at the Atlantic coast, and isolated inland dune occurrences, yield ranges of early Mesozoic to late Archaean ages. Uranium-lead and lutetium-hafnium isotope data provide no evidence of direct derivation of zircon from protosources in crystalline bedrock. The complex distribution patterns can be decomposed into seven 'provenance components' that have previously been encountered in Palaeoproterozoic to Jurassic sedimentary rocks in the region. These components occur in non-random combinations, in proportions that suggest mixing of recycled material from different sedimentary precursors. Taking pre-Cenozoic geology, Cretaceous to Pleistocene drainage history of southern Africa and Holocene wind and current patterns into account, it is possible to work out a consistent model for transport of detritus from sedimentary precursors into the basin, followed by aeolian remobilization within the basin itself. The detrital zircon distributions can thus be interpreted from prior knowledge of the basin filling history, regional geology, geomorphology, drainage evolution and wind pattern. This amounts to a line of reasoning from 'sink back to (intermediate) 'source', which is the opposite of the '(proto)source-to-sink' paradigm commonly invoked in detrital zircon studies.