AbstractAimTo derive null models for the expected number of species shared among multiple samples or habitat patches, allowing exploration of the geometric effects of subdivision on species diversity.LocationGlobal.Major taxa studiedPredominantly sessile organisms.MethodsThe occurrence probability of a species in a subdivided area depends on its abundance and spatial pattern over a known habitat extent. The joint probability that two subareas share a species is the product of the probability of species occurrence in each subarea provided that the latter probability is independent. The sum of this probability over all species is the number of species the two subareas share, or zeta diversity of order 2. Generalizing from 2 tomsubareas yields a null model for zeta diversity of orderm. From zeta diversity, many metrics (e.g., beta and gamma diversity) for themhabitat patches can be calculated, revealing the effects of increasing habitat fragmentation.ResultsThe null models show the geometric effects of subdivision depend on patterns of spatial distribution of species within a landscape and evenness of species abundance distribution. For aggregated assemblages, increasing subdivision decreases shared species, increases beta diversity and results in higher total species richness in subdivided habitat than an equal contiguous area.Main conclusionsTo correctly interpret diversity patterns in fragmented habitat the geometric effects of subdivision must be considered. Our models explain why fragmented habitat could have higher diversity than continuous habitat of equal area but predict a threshold patch‐size above which this will not occur (herec. 100 ha). Apparently positive diversity effects of subdivision, including more species in groups of small patches, are probable outcomes of spatial aggregation of assemblages. The shared species null models offer an analytical tool for exploring the geometric effects of subdivision on diversity while controlling for total habitat area.