Temporal patterns of variability in the longevity of the male and female phases of individual flowers and in the gender expression of plants of a dichogamous New Zealand tree,Pseudowintera colorata (Winteraceae), were documented in field studies. Two measures for the duration of phases in a dichogamous flower are distinguished; the nominal phases based on morphological features of the flower, and the effective phases reflecting the duration of their functions. Flower and phase longevity and phenotypic gender varied considerably throughout the season and among individuals. Temporal variability in phenotypic gender was loosely synchronized among the 12 plants sampled. Three effects of an environmental factor (temperature) were noted. First, increased temperatures shortened the duration of the female phase but had no effect on the duration of the male phase. Second, pollination frequency was positively correlated with temperature. These results indirectly suggest that increased pollination may shorten the duration of the female phase. Third, average population maleness, measured as the proportion of open flowers in the population on a given day which were in the male phase, was positively correlated with temperature. It is postulated that temperature indirectly influences temporal patterns of gender expression in the population through its differential effects on the longevity of the male and female phases in individual flowers. A theoretical model of bet-hedging shows that, if the direction of an environmental effect on the proportions of the sexual phases is irreversible, selection favours asynchronous dichogamy and reduces the temporal variability as much as possible. If the direction of the response is reversible, heterodichogamy is favoured.