Marine and fisheries scientists are increasingly using metapopulation concepts to better understand and model their focal systems. Consequently, they are considering what defines a metapopulation. One perspective on this question emphasizes the importance of extinction probability in local populations. This view probably stems from the focus on extinction in Levins' original metapopulation model, but places unnecessary emphasis on extinction–recolonization dynamics. Metapopulation models with more complex structure than Levins' patch-occupancy model and its variants allow a broader range of population phenomena to be examined, such as changes in population size, age structure and genetic structure. Analyses along these lines are critical in fisheries science, where presence–absence resolution is far too coarse to understand stock dynamics in a meaningful way. These more detailed investigations can, but need not, aim to assess extinction risk or deal with extinction-prone local populations. Therefore, we emphasize the coupling of spatial scales as the defining feature of metapopulations. It is the degree of demographic connectivity that characterizes metapopulations, with the dynamics of local populations strongly dependent upon local demographic processes, but also influenced by a nontrivial element of external replenishment. Therefore, estimating rates of interpopulation exchange must be a research priority. We contrast metapopulations with other spatially structured populations that differ in the degree of local closure of their component populations. We conclude with consideration of the implications of metapopulation structure for spatially explicit management, particularly the design of marine protected area networks.