The index of biotic integrity (IBI) is a commonly used bioassessment tool that integrates abundance and richness measures to assess water quality. In developing IBIs that are both responsive to human disturbance and resistant to natural variability and sampling error, water managers must decide how to weigh information about rare and abundant taxa, which in turn requires an understanding of the sensitivity of indices to rare taxa. Herein, we investigated the influence of rare fish taxa (within the lower 5% of rank abundance curves) on IBI metric and total scores for stream sites in two of Minnesota's major river basins, the St. Croix (n = 293 site visits) and Upper Mississippi (n = 210 site visits). We artificially removed rare taxa from biological samples by (1) separately excluding each individual taxon that fell within the lower 5% of rank abundance curves; (2) simultaneously excluding all taxa that had an abundance of one (singletons) or two (doubletons); and (3) simultaneously excluding all taxa that fell within the lower 5% of rank abundance curves. We then compared IBI metric and total scores before and after removal of rare taxa using the normalized root mean square error (nRMSE) and regression analysis. The difference in IBI metric and total scores increased as more taxa were removed. Moreover, when multiple rare taxa were removed, the nRMSE was related to sample abundance and to total taxa richness, with greater nRMSE observed in samples with a larger number of taxa or sample abundance. Metrics based on relative abundance of fish taxa were less sensitive to the loss of rare taxa, whereas those based on taxa richness were more sensitive, because taxa richness metrics give more weight to rare taxa compared to the relative abundance metrics. ?? 2010 Elsevier Ltd.