In natural systems, organisms are frequently exposed to spatial and temporal variation in predation risk. Prey organisms are known to develop a wide array of plastic defences to avoid being eaten. If inducible plastic defences are costly, prey living under fluctuating predation risk should be strongly selected to develop reversible plastic traits and adjust their defences to the current predation risk. Here, we studied the induction and reversibility of antipredator defences in common frog Rana temporaria tadpoles when confronted with a temporal switch in predation risk by dragonfly larvae. We examined the behaviour and morphology of tadpoles in experimental treatments where predators were added or withdrawn at mid larval development, and compared these to treatments with constant absence or presence of predators. As previous studies have overlooked the effects that developing reversible anti-predator responses could have later in life (e.g. at life history switch points), we also estimated the impact that changes in antipredator responses had on the timing of and size at metamorphosis. In the presence of predators, tadpoles reduced their activity and developed wider bodies, and shorter and wider tails. When predators were removed tadpoles switched their behaviour within one hour to match that found in the constant environments. The morphology matched that in the constant environments in one week after treatment reversal. All these responses were highly symmetrical. Short time lags and symmetrical responses for the induction/reversal of defences suggest that a strategy with fast switches between phenotypes could be favoured in order to maximise growth opportunities even at the potential cost of phenotypic mismatches. We found no costs of developing reversible responses to predators in terms of life-history traits, but a general cost of the induction of the defences for all the individuals experiencing predation risk during some part of the larval development (delayed metamorphosis). More studies examining the reversibility of plastic defences, including other type of costs (e.g. physiological), are needed to better understand the adaptive value of these flexible strategies.