The effect of two training systems (Central Leader with branch pruning versus Centrifugal Training with minimal pruning, i.e., removal of fruiting laterals only) on canopy structure and light interception was analyzed in three architecturally contrasting apple (Malus domestica Borkh.) cultivars: 'Scarletspur Delicious' (Type II); 'Golden Delicious' (Type III); and 'Granny Smith' (Type IV). Trees were 3D-digitized at the shoot scale at the 2004 and 2005 harvests. Shoots were separated according to length (short versus long) and type (fruiting versus vegetative). Leaf area density (LAD) and its relative variance (xi), total leaf area (TLA) and crown volume (V) varied consistently with cultivar. 'Scarletspur Delicious' had higher LAD and xi and lower TLA and V compared with the other cultivars with more open canopies. At the whole-tree scale, training had no effect on structure and light interception parameters (silhouette to total area ratio, STAR; projected leaf area, PLA). At the shoot scale, Centrifugal Training increased STAR values compared with Central Leader. In both training systems, vegetative shoots had higher STAR values than fruiting shoots. However, vegetative and fruiting shoots had similar TLA and PLA in Centrifugal Trained trees, whereas vegetative shoots had higher TLA and PLA than fruiting shoots in Central Leader trees. This unbalanced distribution of leaf area and light interception between shoot types in Central Leader trees partly resulted from the high proportion of long vegetative shoots that developed from latent buds. These shoots developed in the interior shaded zone of the canopy and therefore had low STAR and PLA. In conclusion, training may greatly affect the development and spatial positioning of shoots, which in turn significantly affects light interception by fruiting shoots.