Observational detection of upper stratospheric ozone responses to 27-day solar ultraviolet (UV) variations is often inhibited by larger, dynamically induced ozone variations, which result mainly from the temperature dependence of reaction rates controlling the ozone balance. Here we show that partial correlation coefficients of solar UV and tropical upper stratospheric ozone (1--5 hPa) with the temperature effect removed are larger (0.7--0.8) than are total correlation coefficients of ozone and solar UV (0.4--0.6). The phase lag of ozone relative to solar UV is also increased, and the maximum ozone-UV correlation is obtained at higher altitudes, as compared with correlation analyses using ozone and solar UV data alone. Assuming that temperature variations are not forced by solar UV variations, the ozone sensitivity to solar UV and temperature can be calculated using a linear multiple regression model. The ozone sensitivity to solar UV is generally independent of time periods used for the analysis. However, the magnitude of the ozone sensitivity to temperature at 1--2 hPa increased significantly from solar cycle 21 to solar cycle 22, possibly reflecting long-term changes in the composition of the upper stratosphere. ¿ 2000 American Geophysical Union