Although many aspects of the life history of annual plants have been explained through natural selection for seed set maximization, the predictive power of this concept has rarely been examined. In this dissertation, seed set maximization is used as an optimization criterion in an optimal control model to predict the phenology and carbon allocation schedule for an annual plant, Hemizonia luzulifolia DC. (Asteraceae). This species occurs as two seasonal ecotypes that differ in lifespan and phenology. One has a vernal flowering date and the other an autumnal date. Field measurements of habitat characteristics, growth relations, senescence, and predation on reproductive structures were used as parameters in the optimization model, which visualizes a plant as three compartments: vegetative, reproductive, and storage biomass. The model was then tested to see if it predicts from the relevant habitat differences of the ecotypes the phenological and allocation differences characterizing them. I was particularly interested in whether storage allocation would be a component of the optimal allocation schedule of these or any annuals. Field studies of the two ecotypes also investigated the demographic consequences of their phenological differences, principally the effects on survivorship and seed set. In field comparisons, the autumnal phenology carried a high risk of pre-reproductive mortality, apparently due to variability in the duration of soil moisture levels through the growing season. A vernal flowering date would considerably reduce pre-reproductive mortality in the autumnal population. Autumnal plants that survived the summer drought, however, sometimes had seed sets as much as one hundred times the vernal maximum. Predation on reproductive heads was very low in the vernal population, while in the autumnal population, about thirty percent of the reproductive heads suffered predation. Relative growth rates and senescence rates were similar in the two ecotypes early in the growing season, but relative growth rate dropped during the final few months of the autumnal ecotype's extended lifespan. With constant parameter values, the optimal control model predicted that the autumnal ecotype, but not the vernal, should include carbon storage in its allocation schedule. This resulted from the difference in predation on reproductive heads. When relative growth rate and senescence were considered seasonally varying, the period of carbon storage was considerably longer. The flowering data predictions were qualitatively correct--the autumnal ecotype flowered later than the vernal. The carbon allocation schedules of the two ecotypes were similar. Field plants of both ecotypes apparently stored carbon in stems and roots, but the levels were higher in the autumnal plants. Absolute differences were the result of differences in both biomass and the level of stored carbon (as percent dry weight). This was also observed in plants grown under controlled environments. The