Samples of serum and a series of aqueous standards of SeOp were dried on to the wall of an aluminium cathode cup at a temperature of 293 K, thus ensuring a sample preparation procedure without loss of analyte. Using the Se II 444.62-nm and Se II 444.95-nm spectral lines, the detection limit for the proposed method was established as being 1 ng ml-l. The selenium content of the samples varied between 80 and 120 ng ml-1, which is two orders of magnitude higher than the detection limit of the proposed method. The sodium content of the serum did not interfere in the determination, as the amount of sodium present (140 mmol 1 -1 ) was less than the concentration which can cause a decrease in the spectral line intensity. The relative error of the concentration values is less than 5%. The determination of selenium in serum was performed using direct hollow-cathode discharge atomic emission spectrographic analysis. Optimum conditions for the determination were established. It was found that sputtering of selenium could be achieved, with a satisfactory detection limit and acceptable accuracy, at a power of 240 W. Water cooling was at a flow-rate of 8 I min-1, which provided equal temperatures of the neutral gas and cathode of 430 2 50 K, which is lower than the melting-points of Se and Se02.