The environmental predictability (EP) hypothesis proposes that rapid cold hardening (RCH) might be common in temperate species incapable of surviving freezing events and which also dwell in unpredictable environments. The kelp fly Paractora dreuxi serves as a useful model organism to test this prediction at an intra-specific level because larvae and adults show different responses to low temperature despite occupying a similar unpredictable thermal environment. Here, using acclimation temperatures, which simulated seasonal temperature variation, we find little evidence for RCH in the freeze-intolerant adults but a limited RCH response in freeze-tolerant larvae. In the relatively short-lived adults, survival of -11 degrees C generally did not improve after 2h pre-treatments at -4, -2, 0, 10, 20 or 25 degrees C either in summer- (10 degrees C) or winter (0 degrees C)-acclimated individuals. By contrast, survival of summer-acclimated larvae to -7.6 degrees C was significantly improved by approximately 37% and 30% with -2 and 0 degrees C pre-treatments, respectively. The finding that summer-acclimated larvae showed RCH whereas this was not the case in the winter-acclimated larvae partially supports the predictions of the EP hypothesis. However, the EP hypothesis also predicts that the adults should have demonstrated an RCH response, yet they did not do so. Rather, it seems likely that they avoid stressful environments by behavioural thermoregulation. Differences in responses among the adults and larvae are therefore to some extent predictable from differences in their feeding requirements and behaviour. These results show that further studies of RCH should take into account the way in which differences among life stages influence the interaction between phenotypic plasticity and environmental variability and predictability.