How do we define the limits of a linguistic area? Typologically rare features may spill out beyond the bounds of an otherwise well-defined linguistic area. Rather than viewing the "fuzzy" boundary of a linguistic area as a problem, it can instead be seen to be an integral part of the structure of the linguistic area which may include a core, "depleted core", fringe and even areas beyond the fringe. Clicks are a typical feature of the Kalahari Basin linguistic area, but their patterning on the fringes of this area is not so well-known. Clicks have been borrowed into Bantu languages spoken on the fringes of the area, but their functional load, as measured by the number of click phonemes and frequency of clicks in the lexicon, is lower than in the languages of the core of the area. Clicks have also been borrowed into Bantu languages spoken beyond the fringe of the area, but the functional load of clicks in these ultimate recipients is very low. Processes of click loss, both in Bantu languages and Khoisan languages on the fringe, show the same geographical patterning. The geographical distribution of clicks in southern Africa can be compared to the situation in eastern Africa, where there is evidence for an old linguistic area including Hadza and Sandawe in its core and Dahalo in its fringe.