Abstract. The number of annual weeds were recorded in 752 field experiments in spring-sown cereal crops conducted in Sweden 1972–1993. Two null hypotheses were tested regarding how the sowing date influenced the weed flora. 1. There is no relationship between the weed flora composition and sowing date. A pCCA (with geographic regions, crop species and soil types as covariables) clearly refuted this hypothesis. Hence, the composition of the weed flora varied depending on sowing date. 2. Species classified as summer annuals, winter annuals and germination generalists (that can germinate substantially in both spring and autumn) do not differ in their placement along the first ordination axis in the pCCA, i.e. according to sowing date. An ANOVA was unable to reject this hypothesis. Hence, germination syndrome classification did not explain the observed community differences related to sowing date. These results illustrate the importance of the date of disturbance for any secondary succession involving a seed bank and also the importance of annual dormancy cycles in seed banks.