The niche overlap hypothesis states that "maximal tolerable overlap should vary inversely with the intensity of competition." Rigorous testing of this hypothesis has been complicated by problems of measurement and by the difficulty of formulating operational definitions of "maximal tolerable overlap" and "intensity of competition." We test this hypothesis with direct estimates of niche overlap and competition for the species of a diverse insular small mammal fauna. We recorded multiple captures of six species during 9296 trapnights at 1162 trap stations on Assateague Island (Maryland, USA) during the summer of 1978. To describe the structural niche (microhabitat) of each species, we measured nine habitat variables at each trap station where an animal was captured. Using data on population densities and microhabitats, we estimate pairwise niche overlap and competition with discriminant analysis and multiple linear regression analysis, respectively. We report a positive correlation between niche overlap and competition, rather than the negative relationship predicted by the hypothesis. The negative correlation between niche overlap and species diversity reported for previous tests of the hypothesis may reflect a decrease in the average strength of interaction between species as diversity increases, rather than a competition-mediated decrease in niche overlap.