We used satellite-tracking data from repeated journeys between Europe and West Africa by the same osprey, Pandion haliaetus, individuals to test whether the timing of migration differs between spring and autumn and whether landmarks and stopover goal areas are important for navigation. The timing of migration varied more in autumn than in spring, owing to significant differences between individuals (related to sex) in autumn migration dates. Autumn journeys were significantly slower than spring journeys because they included more stopover days. The difference may be explained by environmental conditions restricting the timing of migration in spring, by differences in opportunities to deposit fuel prior to departure, and by differences in expected changes in foraging/fuelling conditions along the route. Flight paths from repeated journeys by the same individual were often 120-405 km apart (maximum east-west separation 1400 km). These distances exceed the expected normal range of vision, suggesting that the ospreys did not find their way by following familiar landmarks. Flight paths converged in some regions, indicating the existence of up to three intermediary goal areas along the route of individual birds. Between these goal regions route fidelity was low, and the ospreys could find the next goal region after extensive deviation, presumably by map-based navigation and possibly in combination with path integration. (c) 2006 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All tights reserved.