Spirometry was performed in 277 female and 240 male subjects constituting a random, unbiased population sample representative for the population of Denmark. Equipment, calibration and measurement procedures conformed strictly with established standards. Statistical analysis showed that certain power transformations (cubic, quadratic or square root) of the data produced models that fitted the data statistically significantly better than linear models, where no transformations are applied, or logarithmic models where a logarithmic transformation is applied to the response variable. Differences between predictions obtained with the three models were, however, small and without practical significance. Neither rigid adherence to established, well documented standards nor elaborate statistical analysis reduced the inter-individual variance significantly in comparison to previous investigations. The results emphasize that it is important that reference values are appropriate for the methods and populations in question.