During the growing season, some individuals in perennial plant populations may remain alive belowground while others emerge. This phenomenon, known as prolonged dormancy, seems maladaptive, because prolonged dormancy delays growth and reproduction. However, prolonged dormancy may offer the benefit of safety while belowground, leading to the hypothesis that prolonged dormancy is a bet-hedging strategy. We evaluated this hypothesis using a 25-year demographic study of Astragalus scaphoides, an iteroparous perennial plant. First, we determined the relationship between prolonged dormancy and fitness using data from individuals in our population. This analysis showed that prolonged dormancy decreased arithmetic mean fitness and reduced variance in fitness. Geometric mean fitness was maximized at intermediate levels of prolonged dormancy. Empirical patterns of lifetime reproductive success confirm this relationship. We also compared fitness of plants in our population to hypothetical plants without prolonged dormancy, which generally revealed benefits of prolonged dormancy, even if plants could forgo prolonged dormancy without costs to other vital rates. Therefore, prolonged dormancy may indeed function as a bet-hedging strategy, but the benefits of remaining belowground outweigh the costs only for a subset of individuals. Bet hedging has been demonstrated in plants with simple life histories, such as annuals and monocarpic perennials; we present evidence that bet hedging may be important for plants with more complex life histories.