The topic of this paper is the expected (from modelling) and observed sensitivity of the brightness β of noctilucent clouds (NLC) on the ambient water vapour mixing ratio <i>f</i>(H<sub>2</sub>O). Firstly, we show that state-of-the-art models of NLC layer formation predict that in the Arctic summer, a 10% increase of <i>f</i>(H<sub>2</sub>O) in the upper mesosphere should lead to a 22% increase in β. Secondly, we review observations of episodic changes in <i>f</i>(H<sub>2</sub>O) and those in β, the former being available since 1992, the latter since 1979. We also add a new series of observations of <i>f</i>(H<sub>2</sub>O) in the Arctic summer, performed at the ALOMAR observatory (69° N). Thirdly, we show that an increase in daily averaged <i>f</i>(H<sub>2</sub>O) observed in the Arctic summer since 1996, when introduced into the NLC models, comes close to explaining the observed increase in β. In contrast to this gratifying situation for the summer means of <i>f</i>(H<sub>2</sub>O) and β (the latter being available only in summer anyway), the behaviour of annual means of <i>f</i>(H<sub>2</sub>O) is quite different. Those indicate that since 1996 significant decreases of annually averaged upper mesospheric water vapour have occurred at low, mid, and high latitude which can not be used to explain the observed near-stability in NLC brightness over this time period. We close with comments on the very different character of decadal variations in NLC brightness and occurrence rate.