I argue here that, from the perspective of any individual, most landscapes are composed of only three basic types of habitats. These are: (1) source habitat in which reproduction exceeds mortality and the expected per capita growth rate is greater than one; (2) sink habitat, in which limited, reproduction is possible but will not on average, compensate for mortality and the per capita rate of growth is between zero and one; and (3) unusable habitat, which comprises the matrix of all habitats that are never exploited by the species in question, and in which patches of source and sink habitats are embedded. Unlike earlier source-sink models, this model explicitly considers the effects that substituting one type of habitat for another has on the equilibrium size of a population and the interactions between species which can use both source and sink habitats. The model demonstrates that the equilibrium size of a species' population can sometimes be increased by substituting unusable habitat for sink habitat. Thus, even though the average patch quality in the landscape may be decreased, the overall quality of the landscape can increase. For two species with distinct habitat preferences, interactions between species can vary qualitatively as well as quantitatively as a function of the relative abundances of each of the habitat types. The model also shows that the interactions between species are particularly sensitive to the relative costs of moving between patches and sampling patches to determine their quality. Recent fragmentation of natural landscapes may increase the cost of searching for usable (source or sink) patches. Under some conditions, the interspecific interactions may be substantially more negative (competitive) than the interactions that evolved in the original natural landscape, further reducing population sizes and increasing the likelihood of competitive exclusion in fragmented modern landscapes.