Fish species richness is one of the most common descriptors of aquatic biodiversity and an important indicator of the ecological condition of stream reaches and regional biological diversity. However, inadequate sampling effort underestimates the true number of species and, consequently, might provide erroneous information about the ecological conditions of, and trends in, aquatic environments. We sampled 24 to 39 stream reaches (each 150 m long) in seven megadiverse hydrologic units in the Brazilian states of Pará (Amazonia biome) and Minas Gerais (Cerrado biome) to estimate the sampling effort needed to reach asymptotes at hydrologic unit spatial extents. We also estimated how much of the local richness would have been captured if the sampled reach was 50 or 100-m-long. Our results indicate that up to 53 reaches must be sampled per hydrologic unit for attaining an estimated 90% of the fish species richness, depending on the relative number of rare species. In most units, 100-m-long reaches suffice for collecting about 90% of the local sampled richness. We conclude that the effort frequently used in neotropical streams is adequate to capture most of the local fish fauna. However, complete inventories of hydrographic regions would only be feasible with efforts and resources far greater than those commonly employed. Before sampling programs are implemented for estimating true species richness, adequate levels of sampling effort should be assessed wherever that richness is expected to be high.