adequately describes exposure to nitrous oxides in a work environment. The aim of this paper was to assign to the measured data a certain degree of variability which defines the interval in which real concentrations can be found with a given probability. Exposures may be evaluated by using estimates of the geometric mean (GM) and the geometric standard deviation (GSD), i.e. central value and dispersion index, to calculate the confidence interval (Cl) around the mean exposure and compare this interval to the occupational exposure limit. The concentration of nitrous oxides (114 random temporal measurements covering all three shifts, during 6 consecutive days) on coated electrode welding in a car manufacturing plant, was determined with colourimetric direct reading method. Statistical analysis (chi-square and Kolmogorov-Smirnov goodness of fit tests, lognormal and Gaussian distribution fitting and Q-Q plots) was performed. The distribution of the nitrous oxides concentration in the work environment studied closely resembled that of lognormal distribution. The geometric mean was 4.098 mg/m3, median 4.00 Mg/m3, geometric standard deviation 1.829 and 95% confidence interval 3.66-4.58 mg/m3. It is possible to apply the computed GSD for evaluation of exposure limits to nitrous oxides in any other work environment, even with only a few measurements. (J Occup Health 1999; 41: 27-32)