Comparisons of total column ozone measurements from Dobson, Brewer and SAOZ instruments are presented for the period 1990 to 1995 at seven stations covering the mid- and the high northern latitudes, as well as the Antarctic region. The main purpose of these comparisons is to assess, by reference to the well established Dobson network, the accuracy of the zenith-sky visible spectroscopy for the measurement of total ozone. The strengths and present limitations of this latter technique are investigated. As a general result, the different instruments are found to agree within a few percent at all stations, the best agreement being obtained at mid-latitudes. On average, for the mid-latitudes, SAOZ O3 measurements are approximately 2% higher than Dobson ones, with a scatter of about 5%. At higher latitudes, both scatter and systematic deviation tend to increase. In all cases, the relative differences between SAOZ and Dobson or Brewer column ozone are characterised by a significant seasonal signal, the amplitude of which increases from about 2.5% at mid-latitude to a maximum of 7.5% at Faraday, Antarctica. Although it introduces a significant contribution to the seasonality at high latitude, the temperature sensitivity of the O3 absorption coefficients of the Dobson and Brewer instruments is shown to be too small to account for the observed SAOZ/Dobson differences. Except for Faraday, these differences can however be largely reduced if SAOZ AMFs are calculated with realistic climatological profiles of ozone, pressure and temperature. Other sources of uncertainties that might affect the comparison are investigated. Evidence is found that the differences in the air masses sampled by the SAOZ and the other instruments contribute significantly to the scatter, and the impact of the tropospheric clouds on SAOZ measurements is displayed.