A rapid electrothermal atomic-absorption spectrophotometric procedure with homogenization as the only pretreatment is compared with two wet pretreatment methods followed by flame atomic-absorption, in a statistically designed experiment. Samples from the top 5 cm of sediments at three different points of the Tejo estuary were used for the tests. The results show that the electrothermal procedure performed as well as whichever was the better of the flame methods for the determination of cadmium and nickel, irrespective of the sand content of the samples; although not statistically significant, there was apparently a decrease in recovery for chromium, copper and lead by the electrothermal method as the sand content of the sample increased. For zinc the electrothermal method gave results statistically different from those of the flame procedures when the sand content was high but not when the organic content of the sample was high. Because of the limited degree of replication (5 variates) the relative differences between the average values obtained by the three methods were significant only if they exceeded 15%. The electrothermal method has the advantage of substantially greater speed.