The CO2 Slicing analysis of HIRS satellite data for cloud heights at the University of Wisconsin (Wylie and Menzel, 1999) is known to report cloud heights that are below the upper boundary (i.e., the geometric top) of the clouds for semi-transparent cirrus. The reason for the discrepancy between the geometric cloud top and the cloud height derived from the CO2 Slicing analysis is well understood from radiative transfer calculations. It occurs because the CO2 Slicing analysis uses satellite measurements of infrared absorption and emission in clouds. For clouds semi-transparent to terrestrial radiation, absorption and emission occur throughout the cloud so that the CO2 Slicing analysis reports a height indicative of where the absorption and emission occurred inside the cloud. However, some applications need data on the geometric cloud top and need to know the bias of the CO2 Slicing analysis from the geometric top. To statistically evaluate this bias, cloud height data derived from the CO2 Slicing analysis were compared to a more sensitive measurement of the geometric cloud top from the SAGE-II satellite sensor. This study employed 71 clouds where both the HIRS and SAGE-II sensors made height measurements at nearly the same time. The comparison found that the Wisconsin CO2 Slicing analysis underestimated the geometric cloud top by an average of 1.6 km. This bias increased slightly to 2.0 km for upper troposphere clouds. This is close to the previous studies of the CO2 Slicing analysis. ¿ 1999 American Geophysical Union