The stomatal conductance (g) for single leaves and the equivalent canopy conductance for stands of vegetation are often represented in models as empirical functions of saturation vapour pressure deficit or relative humidity. The mechanistic basis of this dependence is very weak. A reanalysis of 52 sets of measurements on 16 species supports the conclusion of Mott & Parkhurst (1991, Plant, Cell and Environment 14, 509–515) that stomata respond to the rate of transpiration (E) rather than to humidity per se. In general, ∂g/∂E is negative and constant so that the relation between g and E can be defined by two parameters: a maximum conductance gm obtained by extrapolation to zero transpiration, and a maximum rate of transpiration Em obtained by extrapolation to zero conductance. Both parameters are shown to be functions of temperature, CO2 concentration, and soil water content. Exceptionally, transpiration rate and conductance may decrease together in very dry air, possibly because of patchy closure of stomata.