In spite of the large number of studies on the role of forests in affecting local and global water and energy cycles, conflicting reports on even the sign of the change in evapotranspiration over forest compared with non-forest land cover can be found depending on the type of data used. Whereas studies based on closure of the water balance suggest higher evapotranspiration over forests, studies based on turbulent exchange and/or energy balance closure suggest generally higher latent heat flux over non-forest sites. In this study, this forest evapotranspiration paradox was investigated using data from four long-term lysimeter stations in western Europe with contrasting land cover conditions. The results were consistent with evapotranspiration estimates from catchment-scale water balance studies rather than with eddy covariance estimates. They were also found to be largely consistent with a model previously proposed to predict forest cover effects on evapotranspiration. The results of this study suggest that eddy covariance data should be treated with care when used to assess long-term average water balance impacts of land use.