Interannual variability of the tropical tropopause is studied using long time series of radiosonde data, together with global tropopause analyses from the National Centers for Environmental Prediction (NCEP) reanalyses over 1957--1997. Comparisons for the period 1979--1997 show the NCEP tropopause temperature is too warm by ~3--5 K and too high in pressure by ~2--6 mbar. However, these biases are approximately constant in time, so that seasonal and interannual variability is reasonably well captured by the NCEP data. Systematic differences in NCEP tropopause statistics are observed between the presatellite (1957--1978) and postsatellite (1979--1997) periods, precluding the use of the reanalyses for the study of multidecadal variability. Interannual anomalies in tropical average radiosonde and NCEP data show variations of order ¿1--2 K over the period 1979--1997, but there can be differences between these two estimates which are of similar magnitude. These differences impact estimates of decadal trends: During 1979--1997, negative trends in tropopause temperature of order -0.5 K/decade are observed in radiosonde data but are not found in NCEP reanalyses. The space-time patterns of several coherent signals are identified in both sets of tropopause statistics. The volcanic eruption of El Chich¿n (1982) warmed the tropical tropopause by ~1--2 K and lowered its altitude by ~200 m for approximately 1--2 years. Smaller tropopause variations are observed following Mount Pinatubo (1991), particularly in radiosonde data. The signatures of the quasibiennial oscillation (QBO) and El-Nino/Southern Oscillation (ENSO) events are strong in tropopause statistics. QBO variations are primarily zonal mean in character, while ENSO events exhibit dipole patterns over Indonesia and the central Pacific Ocean, with small signals for zonal averages. ¿ 2000 American Geophysical Union