The correct identification of colony boundaries is an essential prerequisite for empirical studies of ant behaviour and evolution. Ant colonies function at various organizational levels, and these boundaries may be difficult to assess. Moreover, new complexity can be generated through the presence of spatially discrete subgroups within a more or less genetically homogeneous colony, a situation called polydomy. A colony is polydomous only if individuals (workers and brood) of its constituent nests function as a social and cooperative unit and are regularly interchanged among nests. This condition was previously called polycalic, and the term polydomy was used in a broader sense for a group of daughter nests of the same mother colony (implying limited female dispersal), without regard to whether these different nests continued to exchange individuals. We think that this distinction between ‘polycaly’ and ‘polydomy’ concerns two disparate concepts. We thus prefer the narrower definition of polydomy, which groups individuals that interact socially. Does this new level of organization affect the way in which natural selection acts on social traits? Here, after examining the history of terms, we review all ant species that have been described as expressing polydomous structures. We show that there is no particular syndrome of traits predictably associated with polydomy. We detail the existing theoretical predictions and empirical results on the ecology of polydomy, and the impact of polydomy on social evolution and investment strategies, while carefully distinguishing monogynous from polygynous species. Finally, we propose a methodology for future studies and offer ideas about what remains to be done. © 2007 The Linnean Society of London, Biological Journal of the Linnean Society, 2007, 90, 319–348.