The dryness of the stratosphere has been explained, in general terms, as water condensation from the rising branch of the Hadley cell at the tropical tropopause (Brewer, 1949), but Ellsaesser (1974) suggested that the mean tropical tropopause is not cold enough to account for the observed water vapor mixing ratios. During intense thurderstorms that in part penetrate the tropopause, investigators have observed (1) an increase in local stratospheric water vapor and (2) the temporary presence of air parcels substantially colder than and higher than the tropopause. These cold parcels are calculated to have extremely low water vapor mixing ratios, and their occurrence in the stratosphere suggests a mechanism whereby the effective condensation temperature could be systematically colder than the tropopause. Ice crystals from the cloud, evaporating in the warmer stratosphere, presumably cause the observed increase in water vapor, but mixing of cold desiccated air parcels with lower stratospheric air would tend to decrease its water content. Thus there are opposing factors concerning the role of severe cumulonimbus storms on stratospheric water, and it may require detailed, microphysical analysis to see which effect is larger.