Alfalfa and orchard grass crops were grown at ambient and twice ambient carbon dioxide concentrations in field plots for several years in Beltsville, MD, using semi-open chambers. Canopy conductances throughout many days were determined from water vapour exchange measurements, and indicated significant reductions in canopy conductance to water vapour at elevated carbon dioxide in both species. However, recognizing that the artificial ventilation in the chambers made direct comparisons of evapotranspiration rates questionable, we used a soil–vegetation–atmosphere model to determine what field-scale evapotranspiration rates would have been with natural ventilation. Unlike the ‘omega’ approach, the model used allowed feedbacks between the canopy and the atmosphere, such that, for example, canopy conductance responses affected profiles of temperature and water vapour. Simulations indicated that although canopy conductances were lower at elevated carbon dioxide by as much as 20% in alfalfa and 60% in orchard grass, evapotranspiration rates never differed by more than 3% in alfalfa or 8% in orchard grass. Daily totals of evapotranspiration were only 1–2% lower at elevated carbon dioxide in alfalfa, and 2–5% lower in orchard grass. The results are partly explained by the fact that aerodynamic conductances to water vapour were generally smaller than the stomatal conductance, and also by canopy–atmosphere feedback processes which largely compensated for the lower conductance at elevated carbon dioxide by increasing the gradient for evaporation.