The use of the term ‘metapopulation’ has broadened substantially since its inception to include, for example, subdivided populations that are not necessarily prone to local extinction, and populations with a locally patchy distribution that are not clearly subdivided into discrete demes. This broadened scope has coincided with an exponential increase in the number of articles applying the term (Hanski & Simberloff 1997; Pannell & Charlesworth 2000), but it has arguably come at a cost of precision. In their useful review of the literature on large-scale spatial dynamics in plants, Freckleton & Watkinson (2002) have blown the whistle on this trend. They argue that metapopulations ought only to describe an array of populations (i) that are prone to local extinction, and (ii) that also inhabit discrete and recognizable habitat patches. The first of these criteria is uncontroversial, although the rate of local population extinction that is biologically significant will depend on the nature of the questions being addressed. We believe that the second criterion, however, fails to recognize the utility of the metapopulation approach in studies that are not focused specifically on patch occupancy rates. In this sense, our view thus differs from that of Freckleton & Watkinson (2002), as well as from views expressed recently by Bullock et al . (2002), who also emphasize the importance of fixed habitat patches in defining a metapopulation. It is of course true that many applications of the metapopulation concept in ecology and conservation need to address patch occupancy rates explicitly, and thus require the a priori identification of habitat patches. In a reply to Ehrlen & Eriksson’s (2003) critique of their review, Freckleton & Watkinson (2003) emphasize this point. If we are interested in ensuring the regional conservation of a metapopulation, then efforts must be directed towards the conservation of habitat, whether currently occupied or not. But the dynamics of population turnover in a metapopulation affect not only the regional persistence or survival of a species, but also its population genetics (reviewed in Pannell & Charlesworth 2000) and evolution (reviewed in Ronce & Olivieri 2003) – whether or not we can identify its habitat. Bullock et al . (2002) state that ‘a basic premise of metapopulation theory and models [is] that extinctions make habitat patches available for colonization’ (p. 291). We agree with the implication that an empirical definition of a metapopulation ought to be consistent with its use in the theoretical literature. However, not all theoretical metapopulation models assume the existence of fixed habitat patches. It seems that our differing views stem, at least in part, from a failure to integrate genetic or evolutionary aspects of metapopulation biology with ecological or demographic ones. For example, Freckleton & Watkinson (2002) adopted an entirely demographic perspective in their review and did not consider population genetic structure, because ‘metapopulation theory is not concerned with the movement of genes per se ’ (p. 421). However, the metapopulation perspective has been used fruitfully in both ecology and population genetics, and indeed the concept was first considered by Wright (1940) in the context of population genetics long before Levins (1969, 1970) explicitly introduced the term. Ives & Whitlock (2002) have recently noted that ‘population genetic metapopulations’ may not necessarily equate with ‘ecological metapopulations’, because extinction–colonization dynamics need not affect both the demography and the genetic structure of a species to the same extent. However, because the same underlying processes make the metapopulation concept valuable in population genetics and ecology, consistency in applying the term is desirable.