In June 1996, 16 UV-visible sensors from 11 institutes measured spectra of the zenith sky for more than 10 days. Spectra were analysed in real-time to determine slant column amounts of O3 and NO2. Spectra of Hg lamps and lasers were measured, and the amount of NO2 in a cell was determined by each spectrometer. Some spectra were re-analysed after obvious errors were found. Slant columns were compared in two ways: by examining regression analyses against comparison instruments over the whole range of solar zenith angles; and by taking fractional differences from a comparison instrument at solar zenith angles between 85°and 91°. Regression identified which pairs of instruments were most consistent, and so which could be used as universal comparison instruments. For O3, regression slopes for the whole campaign agreed within 5% for most instruments despite the use of different cross-sections and wavelength intervals, whereas similar agreement was only achieved for NO2 when the same cross-sections and wavelength intervals were used and only one half-day's data was analysed. Mean fractional differences in NO2 from a comparison instrument fall within ± 7% (1-sigma) for most instruments, with standard deviations of the mean differences averaging 4.5%. Mean differences in O3 fall within ± 2.5% (1-sigma) for most instruments, with standard deviations of the mean differences averaging 2%. Measurements of NO2 in the cell had similar agreement to measurements of NO2 in the atmosphere, but for some instruments measurements with cell and atmosphere relative to a comparison instrument disagreed by more than the error bars.