Prolonged seed dormancy in desert annuals is thought to be an adaptation to environmental uncertainty. Germination spread over several years could reduce the year-to-year variation in the fitness of a genotype. Previous work has demonstrated that, for a number of species, not all viable seeds germinate under one set of conditions in the first year, but the subsequent fate of the dormant seeds has not been tested. In this study, germination experiments were performed on seeds of six species of winter annuals from Portal, Arizona. Seeds that do not germinate under good conditions in the first year germinate under the same conditions in subsequent years. However, germination fractions in the first 2 yr are not equal fractions of remaining viable seeds, which is the theoretical optimal behavior, and the germination distributions differ greatly among species. Germination under good conditions following a year of bad conditions for germination also differs among species. Germination behavior is age-dependent, and germination trials conducted on seeds of unknown ages from soil samples may give misleading results. These results indicate that seed dormancy acts as a form of bet hedging but that selection for an optimal distribution of germination across years may be weak.