Recently available isotope ratio infrared spectroscopy can directly measure the isotopic composition of atmospheric water vapour (δ18O, δ2H), overcoming one of the main limitations of isotope ratio mass spectrometry (IRMS) methods. Calibrating these gas-phase instruments requires the vapourisation of liquid standards since primary standards in principle are liquids. Here we test the viability of calibrating a wavelength-scanned cavity ring-down spectroscopy (CRDS) instrument with vapourised liquid standards. We also quantify the dependency of the measured isotope values on the water concentration for a range of isotopic compositions. In both liquid and vapour samples, we found an increase in δ18O and δ2H with water vapour concentration. For δ18O, the slope of this increase was similar for liquid and vapour, with a slight positive relationship with sample δ-value. For δ2H, we found diverging patterns for liquid and vapour samples, with no dependence on δ-value for vapour, but a decreasing slope for liquid samples. We also quantified tubing memory effects to step changes in isotopic composition, avoiding concurrent changes in the water vapour concentration. Dekabon tubing exhibited much stronger, concentration-dependent, memory effects for δ2H than stainless steel or perfluoroalkoxy (PFA) tubing. Direct vapour measurements with CRDS in a controlled experimental chamber agreed well with results obtained from vapour simultaneously collected in cold traps analysed by CRDS and IRMS. We conclude that vapour measurements can be calibrated reliably with liquid standards. We demonstrate how to take the concentration dependencies of the δ-values into account. Copyright © 2010 John Wiley & Sons, Ltd.