In every agroforestry system, the tree canopy reduces the incident radiation for the crop. However, cereal varieties were selected, and most crop growth models were designed for unshaded conditions, so both may be unsuited to agroforestry conditions and performance. In southern France, durum wheat productivity was monitored over 2 years in an agroforestry system including walnut trees and under artificial shade conditions. Yield components were measured in both full and reduced light conditions. The cereal yield was always decreased by shade; by almost 50% for the heaviest shade conditions (31% of light reduction). The main effect of the shade was the reduction in the number of grains per spike (35% at the most) and in the weight of grains (16% at the most). The mean grain weight was moderately affected, while the protein content was increased in shaded con- ditions (by up to 38% for artificial shade). Consequently, the protein yield per hectare was less reduced by the shade than the dry matter grain yield. A crop model (STICS) was also used to simulate the crop productivity in full light and shaded conditions, but the crop LAI and the yield components were not correctly simulated in the shade. The simulations emphasized the sensitivity of the wheat grain filling to shade during the critical period, 30 days before flowering, for yield elaboration. Further experimental and modelling studies should take into account the heterogeneity of shade intensity due to the shape of the tree crown, the width of the crop alley and the orientation of the tree rows and the modifica- tion of carbon allocation inside the plant.