Three improvements to the operational regression procedures used at NOAA/NESDIS to derive total ozone from the TIROS Operational Vertical Sounder (TOVS) measurements are evaluated. An additional predictor, derived from the TOVS measurements, is found to improve the accuracy of the satellite-derived ozone. The substitution for one of the channels used in the regression by another stratospheric channel also improved the accuracy. Finally, the deletion of isothermal cases from the dependent datasets used to derive regression coefficients further improved the accuracy of the ozone determination from the TOVS measurements. The effects of these changes were evaluated with data from the Total Ozone Mapping Spectrometer (TOMS) instrument on Nimbus 7 and the Dobson ozone measurements. Use of these improvements on an independent midwinter set of data from 30° to 65°N resulted in a 4 Dobson Unit (approximately 15%) decrease in the standard deviation of the errors.