We explore relationships among bet-hedging, phenotypic plasticity, and the reliability of environmental cues with respect to the reproductive allocation schedule in annual plants. The optimal schedule of biomass allocation between reproduction and vegetative growth has been shown to be bang-bang in constant environments and graded in unpredictable environments -- a bet-hedging strategy. We summarize genetic-algorithm studies which show that the optimal graded schedule is in fact a weighted combination of the within-season fecundity-maximizing strategies for each of the season-lengths possible in the environment. Each season-length's contribution is a function of its probability, so the degree of optimal gradedness directly reflects the degree of environmental unpredictability. We then develop an analytical model of allocational plasticity which compares fitness between the optimal allocation given an environmental distribution and that given the same distribution but conditioned on the state of an environmental cue. The adaptiveness of plasticity is upper-bounded by an information-theoretic measure of cue reliability: the mutual information of the cue and the environmental state. Simulations confirm that the probability that plastic allocators invade a bet-hedging population increases with cue reliability and that the form of the probability function is that of the mutual-information function. Finally, we discuss the evolution of the complexity of plants' environmental sensitivity and some ecological implications of degrees of perceptual complexity.

Key words: bet-hedging, carbon allocation, environmental perception, genetic algorithms, phenology, phenotypic plasticity