AbstractPremiseDoes the seed size–seed number allocation trade‐off model apply to long‐term persistent soil seed banks? This trade‐off between seed size versus number of seeds produced is usually applied at a single population on an annual basis. Our question is how this model might apply to close relatives that produce dormant seed forming long‐term persistent soil seed banks. These two criteria allow a focus on divergent evolution of conspecifics and permits us to isolate seed size in the spectrum of life history traits that may be influencing seed traits, and on how seed size influences accumulation and persistence in the soil.MethodsIn California, Arctostaphylos species only produce physiologically dormant seed that are fire‐stimulated and that vary in seed size permitting seed size–seed bank density relationship as a test of the seed size–seed number allocation model. Soil seed banks of 10 species of Arctostaphylos were sampled with fruit volumes ranging from 21–1063 mm3. Seed bank density was determined by hand extraction from soil samples.ResultsWe found that seed bank densities were significantly negatively related to fruit or seed size.ConclusionsRather than an issue of allocational trade‐off between size and number, we interpret these results as reflecting seed predation and postfire seedling establishment. Seed bank densities, even after decades, generally were less than one or two‐year's seed production, suggesting intense seed predation. Burial by scatter‐hoarding rodents provided sufficient seeds deep enough for survival of fire. Variation on seed size suggests seedling establishment constraints, but it needs further research.