Attempts to predict the distribution of unionids from readily measurable microhabitat descriptors (i.e. water depth, current velocity, stream size, sediment type) have been largely unsuccessful, but certain biological and calculated hydraulic variables have recently shown some predictive power. We used historic and recent data on unionids (from 1987 to 2003) and hydraulic conditions at 438 sample locations over a 38-km reach of the Upper Mississippi River (Navigation Pool 8) to compare the distribution of unionids with different shell morphologies. We evaluated whether sculptured, thick-shelled (STK) species would be found in areas with higher velocity and shear stress, compared to nonsculptured, thin-shelled (NSTN) species. We used classification trees to model the presence and absence of STK and NSTN species to determine which variables were most likely to predict their distribution. Candidate predictor variables included sampling gear, field substrate, water depth (bathymetry), slope, velocity, shear stress and Froude number under low, moderate and high discharges. Our models predicted that STK mussels would occupy a larger portion of the total aquatic area in this reach of the river than NSTN mussels. However, our data demonstrated that NSTN species used areas of higher shear stress and velocity than STK species, but were also present in backwaters with low energy, thus rejecting our hypothesis. The presence of NSTN species over a wide range of shear stress and velocity was probably due to the wide array of life histories displayed within this guild. Overall, these results are consistent with the flow refuge concept in which unionids are more prevalent in areas with low to moderate hydraulic stresses, regardless of shell morphology, and demonstrate the importance of incorporating abiotic and biotic variables into predictive models.