Troughs of the ocean surface reflect more electromagnetic radiation to a satellite altimeter than crests, causing an apparent lengthening of the altimeter path. This effect, called the 'EM bias,' has been estimated to be a 2--10% of significant wave height (SWH) by various authors from GEOS 3 and aircraft experiments or theoretical analyses. For the case of the SEASAT altimeter, Born et al. (1982) at Jet Propulsion Laboratory (JPL) made an empirical estimate that gave the total correction to the altimeter path length as 7% of the SWH. A large value was expected for SEASAT because an instrumental effect in addition to the EM bias is known to exist. Waveform analyses performed at NASA Wallops flight center (Hayne and Hancock, 1982) give a correction for the instrument effect, but according to the authors their results cannot be compared with results obtained elsewhere for several reasons, including the large uncertainty (30%) of the JPL result. In this paper we resolve the situation for both GEOS and SEASAT altimeters by reinterpreting the results of Hayne and Hancock and making empirical studies using very large data sets to estimate the correction required for the effect of sea state. Empirical estimates of the correction were made by finding a value that minimized (in the sense of least squares) the diferences of many (about 300) repeated pairs of altimeteric profiles for each satellite. The total correction obtained is 1.9% of significant wave height for GEOS 3 and 6.4% for SEASAT, the latter result consistent with the analysis done at JPL of a smaller data set and our interpretation of the Hayne and Hancock waveform analysis.