The seed bank dynamics of five habitats as well as a grazing gradient in the southwestern Kalahari, South Africa were studied. Soil samples were collected in the following habitats: dune crests, dune slopes, dune streets, a calcrete outcrop and a riverbed on the farm Alpha. Soil samples were also collected along a grazing gradient from a watering point. Three methods of soil seed bank analysis were used to analyse the soil samples and to gain insight into soil seed bank response to habitat type and to grazing pressure, over four seasons in the year 2004. Results from the three methods of analysis were also compared to each other. They included the direct seedling germination method, the seedling germination re-examination and the seed extraction method. These analyses were used to (a) estimate seed bank size and composition in response to habitat type and grazing pressure; (b) the differences between the standing vegetation- and the seed bank-flora in different habitats and along a grazing gradient and (c) the type of seed banks that tend to form in certain habitats and in response to grazing pressure. Analyses of soil seed bank size along a grazing gradient showed that the seedling emergence re-examinations estimated a larger size for the seed bank than the direct seedling emergence method. The seed extraction method estimated a significantly larger seed bank size than the other two methods. Heavy grazing pressure favoured annual/opportunistic species such as Schmidtia kalahariensis, which formed very large seed banks in heavily trampled areas. When Schmidtia kalahariensis data was removed from the seed bank analyses, it was found that, in contrast to previous results, the direct germination method mostly estimated a larger seed bank size than the re-examination. Also, the estimation of seed bank size by the flotation method, in this case, was much smaller. The flotation method produced data mostly for hard-seeded species, while the seedling emergence method produced data for species with small seeds and which were readily germinable. In all seasons, the dune crest habitat always had the smallest seed bank and the riverbed habitat always had the largest seed bank. All the dune habitats were characterised by perennial grasses. Perennial grasses formed transient seed banks which were relatively small. The riverbed habitat’s vegetation was mostly composed of annuals. Annual plants formed persistent seed banks which were relatively large. Species richness of the readily germinable seed bank in all habitats, fluctuated between the four seasons and was usually largest in summer. The difference in species richness between the above- and belowground floras fluctuated over four seasons. The dune habitats showed a large difference between the species richness of the above- and the below-ground flora, while the riverbed habitat showed a much smaller difference. The dune habitats had many species with transient seed banks while the riverbed was characterised by many species with short-term persistent and ‘permanent’ seed banks.