In this paper I develop a new viewpoint with regard to species diversity by defining a new type, pattern diversity, to add to the previous categories of inventory and differentiation diversity. By defining and naming this entity, we expand our notions of the properties of ecological units, can integrate it with other ecological concepts, and investigate it. Pattern diversity is a measure of the relative arrangement of subunits within an ecological unit, such as communities in a landscape. In turn, pattern diversity has three types: spatial, temporal, and compositional. Compositional pattern diversity is measured as mosaic diversity using a method called affinity analysis. Mosaic diversity measures landscape complexity and is a function of two properties of species patterns: the variation in species richness among communities and the variation in commonness or rarity among species (evenness). A low value of mosaic diversity is indicative of a simple landscape with one or a few underlying environmental gradients and dominated by a few species; a high value is indicative of a complex landscape with many environmental gradients and no ubiquitous species. Statistical inference can be used for determining standard errors and for comparing with null models. Mosaic diversity is relatively insensitive to the number of sites sampled, the number of species sampled, and the intensity of sampling of rare species. Thus, pattern diversity and its measure–mosaic diversity–hold out promise to provide a new perspective on species diversity and the rules governing the assemblage of communities.