This paper examines the high degree of linear dependence between multiwavelength extinction measurements of stratospheric aerosols and applies this to the problem of separating aerosol and gas contributions to the total measured extinction in the Stratospheric Aerosol and Gas Experiment (SAGE) II and Halogen Occultation Experiment (HALOE) satellite systems. Solar extinction recorded at wavelengths employed primarily for measuring gas densities inevitably includes a contribution from aerosols, whose spectral scattering behaviour depends on their unknown size distribution. Here linear relationships are determined to extract the aerosol contribution from the total extinction at the centers of the SAGE II ozone channel and the HALOE nitrogen dioxide and ozone broadband channels. The derived relationships are applied to both background and volcanic conditions. For the HALOE system we show how the currently employed two-wavelength ratio gives accurate results for most commonly encountered particles sizes in the stratosphere. Although the use of multiple channels can improve the accuracy to which aerosols can be removed from HALOE gas channels, the sensitivity of the relationships to aerosol composition makes them impractical for routine processing. For the SAGE II experiment we show how inaccuracies in aerosol removal have affected the retrieved ozone concentration in the lower stratosphere. Under background conditions, SAGE II ozone concentrations are shown to be biased low by several percent, whereas in postvolcanic periods they are biased high. Prior to the eruption of Mount Pinatubo, SAGE II ozone concentrations below 20 km should be adjusted upward by up to 15% as too much of the measured 0.6 μm extinction has been attributed to aerosol. This change would exacerbate the discrepancy between SAGE II and ozonesonde data in the lower stratosphere throughout the 1980s. In the aftermath of Pinatubo the SAGE II ozone concentrations need to be adjusted downward by up to 15% as too little aerosol extinction was removed from the ozone channel. This shifting bias in the removal of aerosols by the SAGE II retrieval algorithm introduces an apparent downward trend in lower stratospheric ozone concentrations throughout the 1980s and early 1990s as the aerosol size distribution evolves during postvolcanic recovery periods. Trends as large as -2% per year have previously been reported from SAGE II data below 20 km. We show how a significant proportion of this could be explained by the incorrect aerosol removal.¿ 1997 American Geophysical Union