A mechanistic model of carbon partitioning, based on the Munch hypothesis of phloem transport and implemented with PIAF-Munch modelling platform (Lacointe and Minchin 2008), was tested for an architecture more complex than any tested previously. Using (11)C to label photosynthate, responses in transport of photosynthate within a heavily pruned dwarf bean plant (Phaseolus vulgaris L.) to changes in source and sink activities were compared with model predictions. The observed treatment responses were successfully predicted. However, the observations could not be completely explained if the modelled stem contained only one phloem pathway: tracer from a labelled leaf was always detected in both shoot apex and root, whichever of the two leaves was labelled. This shows that bidirectional flow occurred within the stem, with solute moving simultaneously in both directions. Nevertheless, a model architecture with very little more complexity could incorporate such bidirectional flow. We concluded that the model could explain the observations, and that the PIAF-Munch model platform can be expected to describe partitioning in even more complex architectures.