The chemically induced ozone loss inside the Arctic vortex during the winter 1994/95 has been quantified by coordinated launches of over 1000 ozonesondes from 35 stations within the Match 94/95 campaign. Trajectory calculations, which allow diabatic heating or cooling, were used to trigger the balloon launches so that the ozone concentrations in a large number of air parcels are each measured twice a few days apart. The difference in ozone concentration is calculated for each pair and is interpreted as a change caused by chemistry. The data analysis has been carried out for January to March between 370 K and 600 K potential temperature. Ozone loss along these trajectories occurred exclusively during sunlit periods, and the periods of ozone loss coincided with, but slightly lagged, periods where stratospheric temperatures were low enough for polar stratospheric clouds to exist. Two clearly separated periods of ozone loss show up. Ozone loss rates first peaked in late January with a maximum value of 53 ppbv per day (1.6% per day) at 475 K and faster losses higher up. Then, in mid-March ozone loss rates at 475 K reached 34 ppbv per day (1.3% per day), faster losses were observed lower down and no ozone loss was found above 480 K during that period. The ozone loss in hypothetical air parcels with average diabatic descent rates has been integrated to give an accumulated loss through the winter. The most severe depletion of 2.0 ppmv (60%) took place in air that was at 515 K on 1 January and at 450 K on 20 March. Vertical integration over the levels from 370 K to 600 K gives a column loss rate, which reached a maximum value of 2.7 Dobson Units per day in mid-March. The accumulated column loss between 1 January and 31 March was found to be 127 DU (~ 36%).