The along-track scanning radiometer (ATSR) on ERS-1 has delivered a continuous global record of radiometric (skin) sea surface temperature (SST) since August 1991. We present a comprehensive analysis of the large-scale and low-frequency characteristics of the data set using direct comparison with other global SST analyses to develop a quantitative understanding of the various factors contributing to the accuracy of and sources of bias in the first 4 years of the ATSR SST record. Such a global analysis is a necessary complement to direct validation against in situ observations, since large-scale sources of bias may be indistinguishable from instrument noise or sampling uncertainty in individual validation campaigns. No large-scale features attributable to atmospheric contamination through aerosols or water vapor are discernible in the difference between the three-channel, dual-view ATSR SSTs and the National Oceanic and Atmospheric Administration (NOAA) blended analysis of advanced very high resolution radiometer (AVHRR) and in situ SST observations. Features in the difference field can be traced to known deficiencies in the data used in the NOAA operational analysis or tentatively related to skin-bulk temperature differences in certain regions. Atmospheric contamination is, however, evident in the ATSR two-channel SST retrieval (used in daytime and after the failure of the 3.7 μm channel) compared to the NOAA operational analysis. The Pinatubo aerosol plume is the dominant feature of the difference field in the first year of ATSR operation. In all cases, however, the amplitude of the atmospheric signature is significantly lower in dual-view than in corresponding single-view ATSR SSTs, indicating that the potential remains for unbiased two-channel SST retrieval even in the presence of aerosol. ¿ 1998 American Geophysical Union