Dessler (1998) analyzed 60,000 radiosonde profiles to re-examine an analysis done nearly 20 years earlier by Newell and Gould-Stewart (1981) (hereafter: NGS). Contrary to NGS, Dessler finds that the mean tropical tropopause saturation mixing ratio (SMR) is sufficient to explain stratospheric dryness and that there is no need to assume a seasonal or regional preference for water vapor entering the stratosphere. However, in using the SMR to compute the water vapor amount, he implicitly assumes that the air is saturated, since only under this condition are temperature and water vapor physically connected. If the air is not saturated, the average computed by Dessler will overestimate the amount of water crossing the tropical tropopause ([H2O>e), while supersaturation and the presence of ice particles will underestimate [H2O>e. These processes, which are likely to have different regional and temporal distributions, may have fortuitously canceled each other in Dessler's analysis. NGS studied the distribution of tropopause temperatures and focused on the regions and seasons, in which dehydration is more likely to take place. Recent studies indicate that the tropical tropopause has been cooling over the last 25 years and show that the years used by Dessler have the coldest tropical tropopause temperatures. Thus his conclusion may have been different if other years had been studied. These differing viewpoints emphasize the need for a detailed understanding of the stratospheric dehydration mechanism. ¿ 1999 American Geophysical Union