Design of protected areas has focused on setting targets for representation of biodiversity, but often these targets do not include prescriptions as to how large protected areas should be or where they should be located. Principles of island biogeography theory have been applied with some success, but they have limitations. The so-called SLOSS (single large or several small reserves) debate hinged on applications of island biogeography theory to protected areas but was resolved only to the point that parties agreed there might be different approaches in different situations. Although proponents on both sides of the SLOSS debate generally agree that replicating protected areas is desirable, it is difficult to determine how to replicate reserves in terms of number and spatial arrangement. More important, many targets for representation often do not address issues of species persistence. Here, we used a geographic information system in a study of disturbance-sensitive mammals of the Yukon Territory, Canada, to design a protected-areas network that maintains a historical assemblage of species goals for component ecoregions. We simultaneously determined patterns of diversity as Whittaker's beta and compositional turnover and examined how these two measures can give further insights into reserve location and spatial arrangement. Both regional heterogeneity and compositional turnover between nonadjacent sites were significant predictors of the number of protected areas necessary to represent mammals within each ecoregion. Thus, protected-area planners can use diversity measures to identify number and spacing of protected areas within ecologically bounded regions. ©2005 Society for Conservation Biology.