To investigate the hypothesis that disrupting pathogen movement within the plant canopy could slow the development of aschochyta blight, the effect of pea canopy architecture on splash dispersal of Mycosphaerella pinodes-conidia was studied in controlled conditions using a rainfall simulator generating rain events (2 mm) in still air. In intra-plant dispersal experiments, a source constituted by a semi-leafless pea plant with a single infectious lesion (108 pycnidia per cm2 of lesion, 1685 conidia per pycnidium) was placed in the middle of eight healthy target plants. Spore deposition was estimated by the number of lesions that developed on each stipule of the source (auto-deposition) and target (allo-deposition) plants after incubation. Rates of deposition on the source and target plants were 0·53 and 0·47, respectively. On the source plant, most of the spores were splashed downwards, with few spores remaining at the infectious node and very few spores moving upwards. In inter-plant dispersal experiments, potted plants were grouped to constitute 1-m2 canopies. A range of canopy architectures was achieved by using different plant densities and growth stages. A suspension of conidia was placed in the centre of each canopy. Resulting horizontal dispersal gradients were generally described by a negative exponential model. Canopies with a leaf area index (LAI) greater than 0·48 produced gradients with slopes that were not significantly different. A less dense canopy (LAI 0·36) produced a significantly steeper slope. Half-distances were short and ranged between 1·6 and 6·5 cm. The barrier rate, calculated as the ratio of the mean number of lesions assessed on isolated plants to the mean number of lesions assessed on plants in canopies, increased with increasing canopy LAI.