Two ways in which organisms adapt to variable environments are phenotypic plasticity and bet-hedging. Theory suggests that bet-hedging is expected to evolve in unpredictable environments for which reliable cues indicative of future conditions (or season length) are lacking. Alternatively, if reliable cues exist indicating future conditions, organisms will be under selection to produce the most appropriate phenotype – that is, adaptive phenotypic plasticity. Here we experimentally test which of these modes of adaptation are at play in killifish that have evolved an annual life cycle. These fish persist in ephemeral pools that completely dry each season through the production of eggs that can remain in developmental arrest, or diapause, buried in the soil, until the following rainy season. Consistent with diversified bet-hedging (a risk spreading strategy), we demonstrate that the eggs of the annual killifish Nothobranchius furzeri exhibit variation at multiple levels - whether or not different stages of diapause are entered, for how long diapause is entered, and the timing of hatching - and this variation persists after controlling for both genetic and environmental sources of variation. However, we show that phenotypic plasticity is also present in that the proportion of eggs that enter diapause is influenced by environmental factors (temperature and light level) that vary seasonally. In nature there is typically a large parameter zone where environmental cues are somewhat correlated with seasonality, but not perfectly so, such that it may be advantageous to have a combination of both bet-hedging and plasticity.