We examined the effects of excluding rare species on the comparison of species richness. A river benthic data set and a randomization resampling procedure were used to show the importance of this consideration in aquatic bioassessment in particular. The data set was manipulated by deleting species at three levels of rarity as defined by occurrence frequency: once in all 24 replicates from each of three sites, no more than twice, and no more than five times. We focused on differences in species richness because many other bioassessment metrics are dependent on species richness and species composition. Species abundance patterns at the three sites were very different, with many more rare species at the least impacted site than at the more impacted sites. As sample size increased, the diffcrcnces in species richness among the three sites markedly increased in the original data set. However, the exclusion of rare species at the same level of rarity substantially reduced species richness at the least impacted site but had little effect on the most impacted site. This result led to a serious underestimation of the differences in species richness among the sites in terms of both absolute numbers and species loss percentages. Deleting rare species can damage the sensitivity of community-based methods to detect ecological changes; rare species are critical for bioassessment.