Abstract Little is known about the extent to which plant population growth is limited by seed production. We studied two non-sprouting Protea species, both entirely dependent on seeds for recruitment after fire, to determine how flower harvesting would affect the size of the next generation after burning. Five harvesting treatments (0, 25, 50, 75 and 100% inflorescence removal) were applied, each replicated four times in 10 × 10m plots. Seedbanks were censused before a late-summer burn for each species and each replicate. The mean proportion of seeds surviving the burn until germination the following spring was 0.46 for Protea repens and 0.57 for Protea neriifolia. Of the seedlings that emerged, 90% and 55%, respectively, survived the first summer drought. Flower production in the 9 year old stand was strongly related to population density. Optimal plant densities for maximum flower production were estimated as 150 for P. repens and 70 for P. neriifolia. These target densities were greatly exceeded by seedling populations surviving the first summer drought in unharvested stands. If future density-independent mortality is negligible, the excess represents that part of the seedbank that could have been harvested before the burn without influencing future flower production. We developed a simple static model for setting flower harvesting levels in these seed-saturated populations. The predicted harvesting levels (50% of inflorescences in P. repens and 85% in P. neriifolia) were validated against the experimental harvests. We discuss the sensitivity of harvesting levels to variation in fecundity/density relationships, seed survival through a burn and seedling survival until flowering, and the implications for flower harvesting.