Drought control over conductance and assimilation was assessed using eddy flux and meteorological data monitored during four summer periods from 1998 to 2001 above a closed canopy of the Mediterranean evergreen oak tree Quercus ilex. Additional discrete measurements of soil water content and predawn leaf water potential were used to characterize the severity of the drought. Canopy conductance was estimated through the big-leaf approach of PenmanMonteith by inverting latent heat fluxes. The gross primary production (GPP) was estimated by adding ecosystem respiration to net ecosystem exchange. Ecosystem respiration was deduced from night flux when friction velocity (u*) was greater than 0.35ms-1. Empirical equations were identified that related maximal canopy conductance and daily ecosystem GPP to relative soil water content (RWC), the ratio of current soil water content to the field capacity, and to the predawn leaf water potential. Both variables showed a strong decline with soil RWC for values lower than 0.7. The sharpest decline was observed for GPP. The curves reached zero for RWC = 0.41 and 0.45 for conductance and GPP, respectively. When the predawn leaf water potential was used as a surrogate for soil water potential, both variables showed a hyperbolic decline with decreasing water potential. These results were compared with already published literature values obtained at leaf level from the same tree species. Scaling up from the leaf to ecosystem highlighted the limitation of two big-leaf representations: Penman-Monteith and Sellers' II factor. Neither held completely for comparing leaf and canopy fluxes. Tower measurements integrate fluxes from foliage elements dumped at several levels of organization: branch, tree, and ecosystem. The Q. ilex canopy exhibited non-random distribution of foliage, emphasizing the need to take into account a clumping index, the factor necessary to apply the Lambert-Beer law to natural forests. Our results showed that drought is