Variation in dormancy level was tested in seeds of four species, each collected from three populations in 1994 and 1995 (experiment 1). Germination was tested in light and darkness on recently-harvested seeds and on those after-ripened in dry storage for one year. In addition, seeds from each of eight individual plants within each of eight populations were tested for germination when recently harvested and after warm stratification or cold stratification followed by a drying period (experiment 2). Seeds from the two years differed in dormancy level in Silene noctiflora, Sinapis arvensis and Spergula arvensis. Germination percentage differed significantly among populations in Sinapis arvensis and Spergula arvensis in both experiments and in Thlaspi arvense in experiment 2. Furthermore, dormancy level in seeds from different mother plants also varied in the three species tested in experiment 2. Variations at the three levels tested (year, population and mother plant) indicate that these species have a random pattern of variation in dormancy level. It is concluded that variation in seed dormancy among mother plants, populations and years must be taken into account when testing the germination characteristics of a species and also when attempting to model weed seed bank dynamics.