Five Dobson and two Brewer spectrophotometers were used for total ozone observations at Arosa, beginning in 1926 and providing the world's longest series. In this paper we present the results of our attempts to provide a homogeneous series and discuss the data quality problems of the record. From the mid-1950s to 1992, Dobson instrument D15 was calibrated by the statistical Langley plot method. In 1986 the calibration of another Dobson spectrometer at Arosa (D101) was changed by the intercomparison with the primary world Dobson instrument (D83). A statistical model based on simultaneous measurements of D101 and D15 of the period from 1987 to 1990 was used to obtain a total ozone series in line with the primary Dobson spectrophotometer, including a correction for an optical disalignment problem of D15. The series of D101 from 1990 to 1995 was corrected on the basis of data from the Dobson intercomparisons of 1990 and 1995 and comparisons with other total ozone measurements of Brewer and Dobson spectrophotometers at Arosa. A transfer function between Dobson and Brewer spectrophotometric measurements of Arosa is presented, and total ozone measurements of Arosa are compared with version 7 daily overpass data of the satellite instrument the total ozone mapping spectrometer (TOMS) which operated on board Nimbus 7 from autumn 1978 to spring 1993. Available information allowing us to track back the total ozone measurements of Arosa to the measurements of the primary Dobson spectrometer reveal that the total ozone series of Arosa fluctuated no more than approximately 1% against D83 in the period from 1978 to 1995. Average shift of Arosa total ozone data against the TOMS instrument was -1.12 (¿0.1)% over the lifetime of the TOMS instrument. ¿ 1998 American Geophysical Union