This paper reviews the literature on survival estimates for different species of raptors and owls, examines the methods used to obtain the estimates, and draws out some general patterns arising. Estimating survival usually involves the marking of birds so that they can be recognized as individuals on subsequent encounters. Annual survival can then be estimated from: (1) birds ringed at known age (usually as nestlings) and subsequently reported by members of the public (usually as found dead), the ratio of recoveries at different ages being used to calculate annual survival; (2) marked breeding adults, trapped or re-sighted in subsequent years in particular study areas, with the proportion retrapped (or re-sighted) in each year being taken as the minimum annual survival; (3) live encounter (trapped or re-sighted) of birds marked either as nestlings or breeding adults analysed using the capture-mark-recapture (or re-sighting) methods to estimate annual survival; (4) a combination of reports of known-age dead birds and re-trapping/re-sighting of live birds; (5) use of radio-or satellite-tracking to follow the fates of individuals; and (6) the integration of these methods with other information, such as change in numbers between years, to derive estimates of survival and other demographic parameters. Studies confined to particular areas usually give estimates of 'apparent annual survival', because they take no account of birds that leave the area. However, radio-or satellitetracking makes it possible to estimate true survival, including survival of prebreeders that have low natal-site fidelity (this usually requires satellite telemetry). As in other birds, the preferred method for estimating survival has changed over time, as new and more robust methods of estimation have been developed. Methods 1 and 2 were the first to be developed, but without statistical underpinning, while methods 3-6 were developed later on the basis of formal statistical models. This difference has to be borne in mind in comparing older with newer estimates for particular species. Published survival estimates were found for three species of Cathartidae, one of Pandionidae, 29 of Accipitridae, 12 of Falconidae, one of Tytonidae and nine of Strigidae, almost all from temperate Northern Hemisphere species. In most of these species more than one estimate was available, and in some separate estimates for different age or sex groups. The main patterns to emerge included: (1) a significant tendency for annual adult survival to increase with body weight, smaller species having annual survival rates mainly of 60-70%, mediumsized species having rates mainly in the range 70-90% and the largest having rates of > 90%, in the absence of obvious human-caused losses; (2) a lower survival in the first or prebreeding years of life than in subsequent years; (3) a lack of obvious or consistent differences in survival between the sexes, where these could be distinguished; and (4) in