SummaryModels of forest energy, water and carbon cycles assume decreased stomatal conductance with elevated atmospheric CO 2 concentration ([CO 2 ]) based on leaf-scale measurements, a response not directly translatable to canopies. Where canopy-atmosphere are well-coupled, [CO 2 ]-induced structural changes, such as increasing leaf-area index (L D ), may cause, or compensate for, reduced mean canopy stomatal conductance (G S ), keeping transpiration (E C ) and, hence, runoff unaltered.We investigated G S responses to increasing [CO 2 ] of conifer and broadleaved trees in a temperate forest subjected to 17-yr free-air CO 2 enrichment (FACE; + 200 lmol mol À1 ). During the final phase of the experiment, we employed step changes of [CO 2 ] in four elevated-[CO 2 ] plots, separating direct response to changing [CO 2 ] in the leaf-internal air-space from indirect effects of slow changes via leaf hydraulic adjustments and canopy development. Short-term manipulations caused no direct response up to 1.8 9 ambient [CO 2 ], suggesting that the observed long-term 21% reduction of G S was an indirect effect of decreased leaf hydraulic conductance and increased leaf shading. Thus, E C was unaffected by [CO 2 ] because 19% higher canopy L D nullified the effect of leaf hydraulic acclimation on G S .We advocate long-term experiments of duration sufficient for slow responses to manifest, and modifying models predicting forest water, energy and carbon cycles accordingly.