New data for the equilibrium water content of nitrogen, methane and one natural gas mixture are presented. The new binary data and existing binary sets were compared to calculated values of dew point temperature using both the CPA (Cubic-Plus-Association) EoS and the GERG-water EoS. CPA is purely predictive (i.e. all binary interaction parameters are set equal to 0), while GERG-water uses a temperature dependent interaction parameter fitted to published data. The GERG-water model is proposed as an ISO standard for determining the water content of natural gas. The data sets for nitrogen cover the range 233-348 K, and 5-200 bar. Six of these sets, including this work, are well described by both models; five of them have an average dew point temperature deviation of less than 1 K. The seventh set must be rejected, since the data points are too far removed from the other sets. The 13 sets for methane cover the range 233-373 K and 5-249 bar. Seven of these sets have an average bias less than 1 K, while another five have an average deviation less than 2 K. One set must be again rejected, having data too far from the other sets. Two of the remaining sets should probably be rejected as well, since they have large scatter. The data sets that have been measured at low pressures extrapolate well towards the ideal equilibrium values. The two models show similar results, but differ at high pressure and/or temperature. CPA is shown to extrapolate well for methane-water to 1000 bar and 573 K, and our conclusion is that GERG-water must be used with caution outside its specified working range. For some selected natural gas mixtures the two models also perform very much alike. The water content of the mixtures decreases with increasing amount of heavier components, and it seems that both models slightly over-predict the water content of such mixtures. (c) 2007 Elsevier B.V. All rights reserved.