The subantarctic Prince Edward Islands (Marion and Prince Edward) support the largest breeding population of the Vulnerable wandering albatross Diomedea exulans. The numbers of birds breeding in the Marion Island population has fluctuated over the past three decades and appears to be the result of both real changes in the size of the population and changes in the proportion of the population that attempts to breed in a given year. Changes in several demographic parameters that appear to be influenced by both environmental and anthropogenic effects are described. The proportion of first-time breeders in the population was positively correlated with the maximum ENSO index, whereas the annual survival rates of breeding adults were negatively correlated with Japanese pelagic longline fishing effort in the southern Indian Ocean. Adult survival rates were significantly correlated with those recorded on neighbouring Possession Island (Crozet Islands) but differed from those at South Georgia, suggesting common factors operating at an ocean-basin scale. The average survival rate of adult females was lower than that of males. Males who lost partners took 40% longer than females to find a new mate, suggesting a male-biased population. Survival rates of juvenile males and females did not differ. The age distribution of first-time breeders has shifted progressively towards younger birds during the 1990s. Higher than expected survival rates of breeding adults during the late 1990s may be linked to large amounts of supplementary food being made available by the initiation of a longline fishery for Patagonian toothfish Dissostichus eleginoides close to the islands during this time. Overall, breeding success was better than recorded at other localities, indicating that breeding conditions at Marion Island were comparatively favourable. The early implementation of both international and national conservation initiatives to reduce the impact of longline fishing on this species and improve its conservation status is strongly encouraged.