Seed size commonly varies by five to six orders of magnitude among coexisting plant species, a pattern ecologists have long sought to explain. Because seed size trades off with seed number, small-seeded species clearly have the advantage in fecundity, but what is the countervailing advantage of large seeds? Higher competitive ability combined with strong competitive asymmetry can in theory allow coexistence through a competition-colonization trade-off, but empirical evidence is inconsistent with this mechanism. Instead, the key advantage of large seeds appears to be their tolerance of stresses such as shade or drought that are present in some but not all regeneration sites. Here I present a simple, analytically tractable model of species coexistence in heterogeneous habitats through a tolerance-fecundity trade-off. Under this mechanism, the more tolerant species win all of the more stressful regeneration sites and some of those that are less stressful, whereas the more fecund species win most but not all of the less stressful sites. The tolerance-fecundity trade-off enables stable coexistence of large numbers of species in models with and without seed limitation. The tolerance-fecundity mechanism provides an excellent explanation for the maintenance of diversity of seed size within plant communities and also suggests new hypotheses for coexistence in animal and microbial communities.