Until the mid 1970s only a few grey seal Halichoerus grypus pups were born each year on the Isle of May, Scotland, but 1408 pups were born there in 1994. We examined changes in overall site use, individual pupping site fidelity and success in relation to local population density and topography. The area of the island used by seals for breeding has increased as the population increased. Between 1988 and 1994 an index of the proportion of the island's northern area occupied by seals increased from 0.48 to 0.65, while the average population density in these occupied areas decreased over the same period. Although seals have begun to breed in southern parts of the island that were unused previously, other apparently suitable breeding areas on the island remain vacant. Fine scale digital elevation models were constructed using a GIS to evaluate the topographic characteristics of occupied areas and assign topographic ‘costs’ to areas used by seals for breeding on the Isle of May. Seals were associated generally with areas close to the numerous access points from the sea. New areas occupied in the later years of the study had a higher topographical ‘cost’ than the traditional sites. Thus, in an expanding colony such as the Isle of May, areas colonized early in the colony's history were those close to access points and/or standing water and at low elevations. Subsequent expansion resulted in an increase in the areas occupied by seals, but these newly colonized areas were less suitable, and their occupiers were subject to increased topographic ‘costs’. Pup mortality rates were similar in areas of high and low breeding density. Aggressive behaviour between breeding females is the likely mechanism which acts to limit local animal density. Mothers marked at the Isle of May since 1987 returned there to breed with few exceptions, and most females that returned were faithful to their previous pupping sites (median distance between sites used in consecutive breeding seasons = 25 m). Site fidelity persisted even when a previous pupping was unsuccessful and most individuals' pupping locations did not change gradually with time. Comparison with other colonies suggests that the spatial scale of site fidelity is related to the scale of topographic variation within the breeding site.