We describe observations of tropical stratospheric water vapor q that show clear evidence of large-scale upward advection of the signal from annual fluctuations in the effective ''entry mixing ratio'' qE of air entering the tropical stratosphere. In other words, air is ''marked,'' on emergence above the highest cloud tops, like a signal recorded on an upward moving magnetic tape. We define qE as the mean water vapor mixing ratio, at the tropical tropopause, of air that will subsequently rise and enter the stratospheric ''overworld'' at about 400 K. The observations show a systematic phase lag, increasing with altitude, between the annual cycle in qE and the annual cycle in q at higher altitudes. The observed phase lag agrees with the phase lag calculated assuming advection by the transformed Eulerian-mean vertical velocity of a qE crudely estimated from 100-hPa temperatures, which we use as a convenient proxy for tropopause temperatures. The phase agreement confirms the overall robustness of the calculation and strongly supports the tape recorder hypothesis. Establishing a quantitative link between qE and observed tropopause temperatures, however, proves difficult because the process of marking the tape depends subtly on both small- and large-scale processes. The tape speed, or large-scale upward advection speed, has a substantial annual variation and a smaller variation due to the quasi-biennial oscillation, which delays or accelerates the arrival of the signal by a month or two in the middle stratosphere. As the tape moves upward, the signal is attenuated with an e-folding time of about 7 to 9 months between 100 and 50 hPa and about 15 to 18 months between 50 and 20 hPa, constraining possible orders of magnitude both of vertical diffusion Kz and of rates of mixing in from the extratropics. For instance, if there were no mixing in, then Kz would be in the range 0.03--0.09 m2 s-1; this is an upper bound on Kz. ¿ American Geophysical Union 1996