The determination of mercury in urine using cold atomic absorption spectrometry according to Gage and Warren is used after slight modification as standard method for the exposure test in persons exposed to its vapours in industry in Czechoslovakia. Within the concentration range between 100 and 600 nmol X l-1 of urine the relative standard deviation was lower than 4%. In workers exposed to mercury in the production of thermometers, fluctuation of urinary mercury concentration due to variations in diuresis could be greatly reduced by adjustment for endogenous creatinine. After the intake of 2.2 or 2.5 l of liquids the concentration of mercury in urine dropped considerably, but returned again to its initial value within 4 hours after fluid intake. The mercury half-life in the organism is estimated at about 30 days, but may be longer at mercury concentrations of 1 mumol X l-1. Contradictory to the original expectation, the Araki's coefficient b that is recommended for suppressing the effect of the volume of excreted urine on the concentration of urinary substances was found to be concentration-dependent: at 500 nmol X l-1 its value shifted towards 1 (0.85), at 100 nmol X l-1 towards zero (0.2-0.3). It is believed that mercury urinalysis data adjusted for endogenous creatinine are sufficiently reliable to be used for assessing the risk of human exposure to mercury.