The effects of 16 years of continuous browsing by goats in a South African savanna at stocking rates intended for bush control were compared with plots unbrowsed for the same period of time. Differences in bush-clump density, structure and species composition were recorded. Bush-clump density did not differ between browsed and unbrowsed plots. Within individual bush-clumps, browsing was shown to impact more on structure than species composition, with smaller, shorter bush-clumps, containing fewer species but much greater stem-densities. Although species presence/absence was little affected by browsing, many species showed differences in abundance, growth and location within browsed and unbrowsed bush-clumps. Species reduced in abundance in browsed plots included Cussonia spicata, Ehretia rigida, Grewia occidentalis, Jasminum angulare and Senecio linifolius. Several species showed reduced growth in browsed plots, particularly those located at bush-clump edges. The relatively unpreferred Aloe ferox was a notable exception. Although browsing had little effect on the composition of the main clump founding species, emergents or late arrivals, there were twice as many single plants in browsed plots and emergence of several species was restricted to the middle of bush-clumps. Comparison of our findings with aerial photographic evidence and other literature suggest that browsing alone is unlikely to significantly reduce scrub cover, although it can clearly control further expansion. Combinations of fire and browsing, rather than one factor alone, are considered likely to act fastest and most effectively to significantly reduce or remove scrub cover altogether.