The fragmentation of habitat is a major cause of biodiversity loss. However, while numerous studies have suggested that reducing the size of populations and isolating them on fragments leads ultimately to the extinction of a species (small isolated populations are extinction prone), the evidence has been rather conjectural. This is because dispersal is so difficult to measure and isolation difficult to confirm. In past studies, evidence that populations become small and isolated on fragments, leading to declines, has relied on spatial patterns of distribution and abundance. Thus, a species not trapped in the matrix in which fragments are embedded might be assumed isolated on fragments, and if low in abundance on fragments compared to continuous habitat is assumed to have declined on fragments due to this isolation. However, without accurately measuring the degree of isolation, it is difficult to distinguish the role of isolation from other important causes of population decline that are correlated with fragment and population size, such as habitat degradation. Developments in molecular techniques and statistical methods now make it possible to measure isolation. Refreshingly, in this issue Hoehn et al. analyse microsatellite DNA with a suite of statistical methods to show convincingly that a declining species of gecko suffers from greater isolation on habitat fragments than a contrasting gecko that is able to disperse between fragments and hence persist in the severely fragmented wheatbelt of Western Australia.