The concepts of biogeographical regions and areas of endemism are briefly reviewed prior to a discussion of what constitutes a natural biogeographical unit. It is concluded that a natural biogeographical unit comprises a group of endemic species that share a geological history. These natural biogeographical units are termed Wallacean biogeographical units in honour of the biogeographer A.R. Wallace. Models of the geological development of Indonesia and the Philippines are outlined. Areas of endemism within Wallacea are identified by distributional data, and their relationship to each other and to the adjacent continental regions are evaluated using molecular phylogenies from the literature. The boundaries of these areas of endemism are in broad agreement with earlier works, but it is argued that the Tanimbar Islands are biologically part of south Maluku, rather than the Lesser Sundas, and that Timor (plus Savu, Roti, Wetar, Damar, and Babar) and the western Lesser Sundas form areas of endemism in their own right. Wallacean biogeographical units within Wallacea are identified by congruence between areas of endemism and geological history. It is concluded that although Wallacea as a whole is not a natural biogeographical region, neither is it completely artificial as it is formed from a complex of predominantly Australasian exotic fragments linked by geological processes within a complex collision zone. The Philippines are argued to be an integral part of Wallacea, as originally intended. © 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, 101, 193–212.