The resource dispersion hypothesis (RDH), which states that "groups may develop where resources are dispersed such that the smallest economically defensible territory for a pair ... can also sustain additional animals", has been proposed by Macdonald (1983, Nature, Lond.301, 379-384.) as a possible explanation for the formation of social groups in territorial animals. An explicit model was put forward by Carr & Macdonald (1986, Anim. Behav.34, 1540-1549). Our paper analyses their model, and shows that the model predicts that groups larger than a pair, formed in accordance with the RDH, will occur only for a restricted range of parameters. In addition, the independence of territory size and group size, which Carr & Macdonald assumed to follow from their model, only occurs in the particular case of "very rich" patches. A more tractable generalization of the model is briefly discussed. © 1991 Academic Press Limited.