The creation of transgenic plants through genetic engineering has focused interest on how the fitness of a plant species may be altered by small changes in its genome. This study concentrates on a key component of fitness: persistence of seeds overwinter. Seeds of three lines of oilseed rape (Brassica napus subsp. oleifera DC Metzger) and of charlock (Sinapis arvensis L.) were buried in nylon mesh bags at two depths in four habitats in each of three geographically separated sites: Cornwall, Berkshire and Sutherland. Seeds were recovered after 12 and 24 months. Charlock exhibited much greater seed survival (average 60 per cent surviving the first year and 32.5 per cent surviving the second year) than oilseed rape (1.5 per cent surviving the first year and 0.2 per cent surviving the second) at all sites. Charlock showed higher survival at 15 cm burial than 2 cm burial at certain sites, but oilseed rape showed no depth effect. Different genetic lines of oilseed rape displayed different rates of seed survival; non–transgenic rape showed greater survival (2 per cent) than the two transgenic lines, one developed for tolerance to the antibiotic kanamycin (0.3 per cent) and one for tolerance to both kanamycin and the herbicide glufosinate (0.25 per cent). The absolute and relative performances of the different genetic lines of oilseed rape were context specific, illustrating the need to test hypotheses in a wide range of ecological settings.