The total ozone mapping spectrometer (TOMS) ozone measurement is derived by comparing measured backscatter ultraviolet radiances with theoretical radiances computed using standard climatological ozone profiles. Profile shape errors occur in this algorithm at high optical path lengths whenever the actual vertical ozone distribution differs significantly from the standard profile used. These errors are estimated using radiative transfer calculations and measurements of the actual ozone profile. These estimated errors include a short-term component resulting from day-to-day variability in profile shape that gives rise to a standard deviation of 10% in total column ozone amount, as well as a systematic error in the long-term trend at very high solar zenith angles. The trend error resulting from the long-term changes in the ozone profile shape is estimated using measurements from the solar backscattered ultraviolet instrument. At the maximum retrieval solar zenith angle of 88¿, these calculations indicate that TOMS long-term ozone depletions may be overestimated by 5% per decade. For trend studies that are restricted to latitudes lower than 60¿ (a maximum of 83¿ solar zenith angle), this error is reduced to no more than 1--2% per decade. Differential impact of the profile shape error at the various TOMS wavelength pairs indicates that profile shape information is present in the TOMS measurements at high solar zenith angles. An interpolation method internal to TOMS is proposed to extract this information. It improves the retrieval at high solar zenith angle, reducing the short-term variability to a standard deviation of 5%, and essentially eliminates the long-term error. The set of standard profiles used in the algorithm are adjusted based on an analysis of empirical orthogonal functions derived from a composite climatology of Stratospheric Aerosol and Gas Experiment II and balloonsonde profile measurements.¿ 1997 American Geophysical Union