Data most commonly used at present to calibrate measurements of mercury vapor concentrations in air come from a relationship known as the "Dumarey equation". It uses a fitting relationship to experimental results obtained nearly 30 years ago. The way these results relate to the international system of units (SI) is not known. This has caused difficulties for the specification and enforcement of limit values for mercury concentrations in air and in emissions to air as part of national or international legislation. Furthermore, there is a significant discrepancy (around 7% at room temperature) between the Dumarey data and data calculated from results of mercury vapor pressure measurements in the presence of only liquid mercury. As an attempt to solve some of these problems, a new measurement procedure is described for SI traceable results of gaseous Hg concentrations at saturation in milliliter samples of air. The aim was to propose a scheme as immune as possible to analytical biases. It was based on isotope dilution (ID) in the liquid phase with the (202)Hg enriched certified reference material ERM-AE640 and measurements of the mercury isotope ratios in ID blends, subsequent to a cold vapor generation step, by inductively coupled plasma mass spectrometry. The process developed involved a combination of interconnected valves and syringes operated by computer controlled pumps and ensured continuity under closed circuit conditions from the air sampling stage onward. Quantitative trapping of the gaseous mercury in the liquid phase was achieved with 11.5 μM KMnO4 in 2% HNO3. Mass concentrations at saturation found from five measurements under room temperature conditions were significantly higher (5.8% on average) than data calculated from the Dumarey equation, but in agreement (-1.2% lower on average) with data based on mercury vapor pressure measurement results. Relative expanded combined uncertainties were estimated following a model based approach. They ranged from 2.2% to 2.8% (k = 2). The volume of air samples was traceable to the kilogram via weighing of water for the calibration of the sampling syringe. Procedural blanks represented on average less than 0.1% of the mass of Hg present in 7.4 cm(3) of air, and correcting for these blanks was not an important source of uncertainty.