While theoretical studies predict that inducible defences should be fine-tuned according to the qualities of the predator, very few studies have investigated how dangerousness of predators, i.e. the rate at which predators kill prey individuals, affects the strength of phenotypic responses and resulting benefits and costs of induced defences. We performed a comprehensive study on fitness consequences of predator-induced responses by involving four predators (leech, water scorpion, dragonfly larva and newt), evaluating costs and benefits of responses, testing differences in dangerousness between predators and measuring responses in several life history traits of prey. We raised Rana dalmatina tadpoles in the presence of free-ranging predators, in the presence of caged predators, and exposed naive and experienced tadpoles to free-ranging predators. Tadpoles adjusted the intensities of their behavioural and morphological defences to predator dangerousness. Survival was lower in the nonlethal presence of the most dangerous predator, while we could not detect costs of induced defences at or after metamorphosis. When exposed to free-ranging predators, small, but not large, tadpoles benefited from exhibiting an induced phenotype in terms of elevated survival when compared to naive tadpoles, but we did not observe higher survival either in tadpoles exhibiting more extreme phenotypes or in tadpoles exposed to the type of predator they were raised with. These results indicate that while predator-induced defences can mirror dangerousness of predators, costs and benefits do not necessarily scale to the magnitude of plastic responses.