Estimating the frequency of long-distance pollination is important in cultivated species, particularly to assess the risk of gene transfer following the release of genetically modified crops. For this purpose, we estimated the diversity and origin of fertilizing pollen in a 10 × 10 km French oilseed rape production area. First, the cultivar grown in each field was identified through surveys to farmers and using microsatellite markers. Examination of the seed set in fields indicated high rates of seed contamination (8.7%) and pollination from other sources (5%). Then, male-sterile plants were scattered over the study area and their seed genotyped using the same markers. Most pollination was local: 65% of the seeds had a compatible sire in the closest field, i.e. at 50 or 300 m depending on site, but the nearest compatible field was found more than 1000 m away for 13% of the seeds. To assess the diversity of fertilizing pollen, each seed was assigned to the nearest putative siring cultivar. The observed diversity of pollen was then compared to that predicted by simulations using three empirical dispersal models with increasing proportion of long-distance pollination. The diversity was sensitive to the dispersal kernel used in the simulations, fatter-tailed functions predicting higher diversities. The dispersal kernel that was more consistent with our data predicted more long-distance dispersal than the exponential function.