Empirical research has for a long time observed that animal densities may both increase and decrease with patch size, and these variable responses have been difficult to explain using the current theoretical framework. The most influential hypothesis, the resource concentration hypothesis, predicts only positive density–area relations, as a consequence of different emigration and immigration rates in small and large patches, and empirical deviations have inspired a flurry of alternative explanations. In this paper, we use realistic rules for the relationship between patch size and migration rates and show a wider predictive range for density–area relations than previously believed. Comparisons with published data suggest that observed density–area relations may easily fit in a framework based on a minimum set of behavioural and population processes. This does not imply that other mechanisms are unimportant, but merely that their quantitative importance can only be evaluated relative to patch geometry and local growth.