During a 2-week intercomparison of ground-based zenith sky visible spectrometers in September 1994 at Camborne, United Kingdom (50 ¿N, 5 ¿W), ozone profiles were measured by electrochemical cell (ECC) sondes during 11 twilight periods. We use these profiles and a radiative transfer model to calculate separate air mass factors (AMFs) for each twilight period. We examine ozone data from one of the spectrometers of the Syst¿me d'Analyse par Observation Z¿nithale (SAOZ) design and we show that these separate AMFs give very straight Langley plots, except at solar zenith angles exceeding 90¿. Total ozone calculated using these AMFs, by a variety of commonly used procedures, agrees with the total ozone calculated by vertically integrating the sonde profiles, with mean differences of 0 to 7 Dobson units (DU), depending on the method, and standard deviations of 9 to 11 DU (1 &sgr;). Total ozone calculated using the best procedure (i.e., averaging twilight values), which is not sensitive to errors in the gradients of AMFs, gave excellent agreement whether using separate AMFs or fixed climatological AMFs. We analyse the variance of the data set and several sources of systematic error in the measurements. We also illustrate from an example during the campaign that such analyses are pointless in the presence of a strong jet stream, which can give rise to changes in ozone during the course of the day that are large enough to invalidate the Langley plot.¿ 1997 American Geophysical Union