The development of the ''split window'' approach for correcting satellite measurement of radiance for atmospheric attenuation is reviewed. Then the theoretical results are compared to results from actual measurements which consist of satellite measurements in the three infrared windows of the AVHRR. Ground truth for the comparisons comes from buoys. The satellite measurements were screened for clouds, and the remaining ones were used in the analysis. Using the data set, several statistical analyses were performed. These showed that, when the two channels that are truly a split window are used, the result of the statistical model agrees with the one derived from theoretical considerations. When the 3.8-μm channel is combined with one in the 10--12 μm region, the result of the statistical model does not take the split window form. Results show that the method is capable of producing sea suface temperatures with a standard deviation of 1 K or less.