Low-frequency variability in global mean sea level (GMSL) is studied for the period 1950--2000 by interpolating sparse tide gauge data to a global grid using empirical orthogonal functions (EOFs) of sea level variability determined from TOPEX/Poseidon (T/P) altimeter data. Results are based on data with long-term trends removed. The fact that the results do not have secular trends is an artifact of the analysis and should not be interpreted as an indication that sea level is not rising. The EOF reconstruction technique is discussed, and the resulting GMSL time series is compared to GMSL time series from Geosat and T/P altimetry and proxy GMSL time series estimated from global sea surface temperature data. The error assessment suggests the accuracy of the GMSL time series reconstructed from the tide gauge data is 2--4 mm RMS for a 1 year running mean smoothing and about 1 mm for a 5 year running mean smoothing. Several El Ni¿o/La Ni¿a events are evident in the GMSL, as well as significant low-frequency variability at a 10--12 year period. GMSL appears to have been generally lower than normal in the late 1960s, throughout the 1970s, and in the 1980s. GMSL appears to have been generally higher than normal in the late 1950s and early 1960s and in the early 1980s and has been rising throughout the 1990s, when T/P is observing. The implication of the low-frequency signals on the determination the secular rate of GMSL from satellite altimetry is discussed.