1. Currently, there is no general agreement about the extent to which predators impact prey population dynamics and it is often poorly predicted by predation rates and species abundances. This could, in part be caused by variation in the type of selective predation occurring. Notably, if predation is selective on categories of individuals that contribute little to future generations, it may moderate the impact of predation on prey population dynamics. However, despite its prevalence, selective predation has seldom been studied in this context. 2. Using recoveries of ringed tawny owls (Strix aluco) predated by 'superpredators', northern goshawks (Accipiter gentilis) as they colonized the area, we investigated the extent to which predation was sex and age-selective. Predation of juvenile owls was disproportionately high. Amongst adults, predation was strongly biased towards females and predation risk appeared to increase with age. This implies age-selective predation may shape the decline in survival with age, observed in tawny owls. 3. To determine whether selective predation can modulate the overall impact of predation, age-based population matrix models were used to simulate the impact of five different patterns of age-selective predation, including the pattern actually observed in the study site. The overall impact on owl population size varied by up to 50%, depending on the pattern of selective predation. The simulation of the observed pattern of predation had a relatively small impact on population size, close to the least harmful scenario, predation on juveniles only. 4. The actual changes in owl population size and structure observed during goshawk colonization were also analysed. Owl population size and immigration were unrelated to goshawk abundance. However, goshawk abundance appeared to interact with owl food availability to have a delayed effect on recruitment into the population. 5. This study provides strong evidence to suggest that predation of other predators is both age and sex-selective and that selective predation of individuals with a low reproductive value may mitigate the overall impact of predators on prey population dynamics. Consequently, our results highlight how accounting for the type of selective predation occurring is likely to improve future predictions of the overall impact of predation.