(1) Edaphically-restricted rare species differ in some aspects from closely related, widespread congeners, but not in others. Characteristics that do not differ between three rare mariposa lilies (Calochortus obispoensis Lemm., C. tiburonensis Hill, and C. pulchellus Dougl. ex Benth.) and one common species (C. albus Dougl. ex Benth.) include mean number of seeds set per capsule, mean seed weight, germination behaviour and dormancy mechanisms. (2) Differences between populations of two rare species (C. obispoensis and C. tiburonensis) and the common species, that favour the persistence and proliferation of the common species, are: (i) minimum bulb size for reproduction; (ii) mean number of flowers and fruits produced, grazed and fruit set; (iii) the extent of bulbil/bulblet production. However, one rare species (C. pulchellus) appears at least as robust sexually as the common species, and may only be rare because of its limited geographic distribution. Year-to-year changes in reproductive characteristics in the four species appear unpredictable. (3) In some years, populations of three rare species in this genus at least equal in sexual reproduction that of a common congener. This episodic behaviour may ensure the persistence of rare species, despite long periods of low reproductive output. (4) The rare C. obispoensis and C. tiburonensis have low seed survival and seedling establishment, low adult mortality and slow growth. The same pattern is true for the third rare species, C. pulchellus, but it has two size-classes of reproductive individuals. The common C. albus behaves much like the rare C. pulchellus, but differs from all rare species in its much greater seed output because of its greater probability of survival to reproductive size. (5) The rare species may be considered climax species, with populations near equilibrium in unusual soil environments. The more common species has life-history characteristics of a colonizer. (6) Rarity in populations of three mariposa lilies appears rather idiosyncratic, and few generalizations should be drawn across taxa or conspecific populations.