The aim of this study was to develop a method for selenium determination in urine and examine the influence of sensitivity enhancement reagents, instrument parameters, internal standards and the salt content of the urine matrix on the determination. Several carbon-containing solutions (methanol, ethanol, propanol, butanol, glycerol, acetonitrile and acetic acid) were examined for their sensitivity enhancement effect. Enhancement factors up to six were obtained and were dependent on the nebulizer gas flow and rf power. There was no important difference in the enhancement effects of these solutes when the nebulizer gas flow rate was optimized for each solute. The influences of sample uptake rate, nebulizer flow rate and rf power were examined in multivariate experiments. The nebulizer gas flow rate and rf power were found to be interdependent, but the sample pump flow rate was independent of the other parameters. The sensitivities of different selenium species (selenite, selenate, selenomethionine and trimethylselenonium iodide) were equal during the experiments in different enhancement solutes and when analysed with the optimized parameter settings. The influence of the urine matrix and different salts on four selenium isotopes were examined. It was concluded that only Se-82 was usable for quantitative determination in urine as the blank at mass 82 was close to zero. The blank values at masses 76, 77 and 78 varied considerably and differently with different salts and salt concentrations. Yttrium,; indium and gallium were compared for their use as internal standards. There was no difference in the efficiency of these internal standards and they were all usable. The accuracy of the method was determined using the NIST SRM 2670 Toxic Metals in Freeze-dried Urine (certified value 0.030 +/- 0.008 mg l(-1)). The result was 29.4 +/- 1.0 mu g l(-1) (n = 8) and the precision was 3.4%. In synthetic urine diluted 1 + 1, the limit of detection was 0.9 mu g l(-1) and the limit of determination was 2.2 mu g l(-1)