Different authors disagree on whether human genome variation should be described as continuous or discontinuous; in the latter case, by attributing an individual's genotype to one genetic cluster, one would also obtain information on the individual's genome in general. An analysis of 377 microsatellites of the CEPH human diversity panel was interpreted as evidence that most genotypes cluster into one of five distinct groups, approximately corresponding to continents, which were pro- posed by some authors as the major biological subdivisions of humankind. Here we analyse the same dataset by a specific numerical method, designed to detect genomic boundaries, i.e. zones of increased change in maps of genomic variation. We show that statistically significant boundaries can be described between groups of populations, but different clusters are identified, depending on the assumptions of the model. In addition, these clusters do not correspond to the clusters inferred from previous analyses of the same or of other polymorphisms. We conclude that it is indeed possible to cluster genotypes according to geography, but no study so far identified unambiguously anything that can be regarded as a major genetic subdivision of humankind, and hence discontinuous models of human diversity are unsupported by data.