The error budget of the Stratospheric Aerosol and Gas Experiment (SAGE) II ozone profile measurements is discussed in depth. Five ozone profiles are compared against coincident ROCOZ-A and electrochemical concentration cell (ECC) ozonesonde measurements at Natal, Brazil (6¿S) and Wallops Island, Virginia (36¿N). The mean difference between all the measurements is approximately 1% and the agreement is within 7% at all altitudes between 20 and 53 km. Using datasonde and National Weather Service satellite observations of temperature, the agreement is almost equally as good for ozone mixing ratios on pressure surfaces. A comparison of the intrinsic SAGE II measurement errors with measured tropical variances suggests that the precision of the ozone profiles is approximately 5% between 24 and 36 km, degrading to 7% at an altitude of 48 km. It is inferred that the measurement errors possess a vertical correlation distance of 3 km and that therefore the profile precision improves by a factor of approximately 1.3 if the profiles were to be smoothed over 5 km vertically. The repeatability of SAGE II ozone profiles depends also upon the uncertainty in the profiles' reference altitudes. These errors are correlated over 7-day periods and increase to 200 m over that period of time. The accuracy of SAGE II ozone profiles is expected to be 6% above 25 km altitude. The SAGE II profiles provide useful ozone information up to approximately 60 km altitude and are more precise than the SAGE I profiles. SAGE II profiles, combined with revised SAGE I profiles, form an excellent data base for estimating the long-term trend in stratospheric ozone since 1979. ¿ American Geophysical Union 1989