A system to determine the spectral responsivity of ultraviolet (UV) radiometers has been developed and is routinely operated at the Central Ultraviolet Calibration Facility, at the National Oceanic and Atmospheric Administration. The instrument and the measurement methodologies are described. Results of measurements from thermally controlled broadband UV radiometers of the Robertson-Berger (R-B)-type are described. Systematic differences in the spectral response curves for these instruments have been detected. The effect of these differences on the field operation of UV-B radiometers has been studied by calculating the instrumental response from modeled UV spectra. The differences of the weighted spectral UV irradiances, measured by two radiometers with different spectral response functions, caused by the daily variation in the position of the sun were estimated for fixed values of total ozone, altitude and albedo, and for cloud-free conditions. These differences increase with the solar zenith angle and are as large as 8%. Larger differences in the instrumental response may be produced by ozone variations. Thus, care must be taken when analyzing data from R-B radiometers and comparing results from different instruments. Routine cycling of UV-B radiometers in operative networks without a careful determination of the spectral responsivity, or small drifts of the spectral responsivity, may strongly affect the accuracy of UV radiation measurements and produce an erroneous trend. Because of the possible differences among radiometers, it would not be practical to derive the long-term behavior of UV radiation without routine and thorough characterization of the spectral responsivities of the instruments.