This paper describes a stratospheric density and temperature retrieval experiment based on the solar occultation measurement of the Stratospheric Aerosol and Gas Experiment (SAGE II). The entire retrieval analysis involves two inversion steps: the vertical structure inversion, which derives the profile of local atmospheric extinction from SAGE II limb optical depth data, and the species inversion, which inverts the concentration of air molecules, aerosols, ozone, and nitrogen dioxide from the derived atmospheric extinction at five SAGE II short wavelengths (0.385, 0.448, 0.453, 0,525, and 0.600 μm). The inversion schemes for these steps are Twomey's (1975) modification of Chahine's nonlinear inversion algorithm and the Levenberg‐Marquardt nonlinear least squares method, respectively. The derived density profile is then used to infer the temperature distribution, assuming that the atmosphere is in hydrostatic equilibrium and obeys the ideal gas law. The retrieval analysis is illustrated in detail by using a SAGE II measurement event that occurred on September 15, 1987 (43.6°N, 4.1°W). An estimate of the error associated with the retrieved molecular density and temperature profiles based on the propagation of the instrument measurement error, the reference altitude error, and an assigned 10% error to the pressure at the top measurement level is also provided. The temperature profiles retrieved from the SAGE II observations are compared with near‐coincident, in both time and space, French Rayleigh lidar and NASA Wallops Flight Facility rocket datasonde soundings as well as the National Meteorological Center (NMC) data analyses. The results indicate that the mean SAGE II temperature agrees with the mean lidar measurement to within 2°C at altitudes from 30.5 to 52.5 km. The SAGE II and datasonde observations agree to within about 4°C in approximately the same altitude region. The mean temperature difference between the SAGE II and NMC analyses are, on average, about 1°, 4°, 2°, and 3°C at the standard pressure levels of 10, 5, 2, and 1 mbar,