The Dobson spectrophotometer is the primary standard instrument for ground-based measurements of total column ozone. The accuracy of its data depends on the knowledge of ozone absorption coefficients used for data reduction. We document an error in the calculations that led to the set of absorption coefficients currently recommended by the World Meteorological Organisation (WMO). This error has little effect because an empirical adjustment was applied to the original calculations before the coefficients were adopted by WMO. We provide evidence that this adjustment was physically sound. The coefficients recommended by WMO are applied in the Dobson network without correction for the temperature dependence of the ozone absorption cross sections. On the basis of data measured by Dobson numbers 80 and 82, which were operated by the National Oceanic and Atmospheric Administration (NOAA) Climate Monitoring and Diagnostics Laboratory at the South Pole, we find that omission of temperature corrections may lead to systematic errors in Dobson ozone data of up to 4%. The standard Dobson ozone retrieval method further assumes that the ozone layer is located at a fixed height. This approximation leads to errors in air mass calculations, which are particularly relevant at high latitudes where ozone measurements are performed at large solar zenith angles (SZA). At the South Pole, systematic errors caused by this approximation may exceed 2% for SZAs larger than 80¿. The bias is largest when the vertical ozone distribution is distorted by the ozone hole and may lead to underestimation of total ozone by 4% at SZA = 85¿ (air mass 9). Dobson measurements at the South Pole were compared with ozone data from a collocated SUV-100 UV spectroradiometer and Version 8 overpass data from NASA's Total Ozone Mapping Spectrometer (TOMS). Uncorrected Dobson ozone values tend to be lower than data from the two other instruments when total ozone is below 170 Dobson units or SZAs are larger than 80¿. When Dobson measurements are corrected for the temperature dependence of the ozone absorption cross section and accurate air mass calculations are implemented, data from the three instruments agree with each other to within ¿2% on average and show no significant dependence on SZA or total ozone.