In agroforestry systems (AFS) with annual intercrops, the competition for light, water and nitrogen plays a key role. Mature trees capture a larger part of the resources, and the yield of the intercrop decreases. When the tree stem density is high (>50 stems.ha -1), the yield of the intercrop may therefore drop to unprofitable levels. A possible solution to that competition problem would be to reduce the size of the trees by pollarding. Pollarded trees produce both timber with the bole and biomass or fodder with the leaves and branches that are regularly pruned every 2 to 10 year. The reduction of the size of the crown would also be profitable to the intercrops, allowing a sustained and profitable yield even with old trees. The concept was evaluated both by modeling with the Hi-sAFe numerical model and by an experiment comparing standard walnut trees and pollarded walnut trees. Numerical runs of the model indicate that AFS with pollarded trees may reach very high levels of tree-crop complementary, achieving higher Land Equivalent Ratios than AFS with standard trees. However, the model predicts that the diameter growth of the trunk could be heavily reduced by the pollarding regime. These predictions should be confirmed by field experiment. We settled an experiment with walnut trees aged 16 years in spring 2013. During the first growing season after pollarding, the girth growth of the pollarded trees was surprisingly not reduced, as compared to standard unpollarded trees. Some hypotheses to explain this behaviour are discussed. If this result is confirmed during the next years, the model should be modified to incorporate this aspect, and the productivity of the AFS with pollarded trees could be even higher.