The extrapolation of no observed effect concentrations derived from experimental ecosystem tests (mesocosms) to the field assumes low assessment factors, i.e., the test system is sensitive and serves as a good surrogate for the field. We tested this assumption by comparing the responses of dominant invertebrate taxa exposed to municipal effluent in Procter&Gamble's experimental stream facility (ESF) to those of the same taxa found in routine field samples collected by the Ohio Environmental Protection Agency (1,112 sites). The index of community sensitivity was used as the basis for comparison of mesocosm responses to field-collection data. Control ESF responses corresponded to the 75th and 80th percentiles, whereas most effluent responses corresponded to between the 22nd and 30th percentiles. These analyses support the use of a low assessment factor for the ESF.