Severe bottlenecks can reduce genetic diversity and increase inbreeding as individuals are forced to mate with close relatives, but it is unknown at what minimum population size the negative fitness consequences of bottlenecks are expressed. The New Zealand avifauna contains a large number of species that have gone through bottlenecks of varying severity, providing an exceptional opportunity to test this question by using the comparative method. Using decreased hatchability as a measure of fitness costs, we found that hatching failure was significantly greater among both native and introduced species that had passed through bottlenecks of <150 individuals. Comparisons between pre-and postbottleneck populations of introduced species suggest that hatching problems arise even in populations founded by <600 individuals. Our study confirms that hatching failure is widespread and persistent among birds passing through severe bottlenecks and that the population sizes at which this fitness cost is expressed are several times greater than the number of individuals currently used to found most new populations of endangered species. We recommend that conservation managers revise the protocols they use for reintroductions or they may unwittingly reduce the longterm viability of the species they are trying to save. H abitat destruction and exploitation are causing catastrophic population declines in many species around the world. Even if endangered populations recover, severe bottlenecks may reduce genetic diversity and increase inbreeding as survivors are forced to mate with close relatives, resulting in lowered heterozygosity, increased genetic load, and increased expression of deleterious alleles (1). Inbreeding may yield significant costs to fitness and decrease population survival (2), a process termed inbreeding depression, but its importance has been questioned (3-5), and examples of the negative fitness consequences due to inbreeding in small populations of wild animals are few (6, 7). Despite the potential importance of bottleneck size to conservation biology, the number of individuals required to avoid the fitness costs of small population size and maintain the viability of a population has been difficult to test in free-living animals (1).Theoretical models suggest that minimum effective population sizes range from 50 to 5,000 individuals, depending on levels of acceptable loss of genetic variability and the timeframe of population persistence (8, 9). The exact number is not a trivial question because the survival of many endangered species depends on the reliability of such guidelines. It has even been suggested that severe bottlenecks may be advantageous because they reduce inbreeding depression by purging deleterious alleles (10) although whether such benefits are great enough to justify deliberate inbreeding have been questioned (11,12). The problem for conservation biologists is to understand whether bottlenecks create fitness costs and at what population size these costs become so severe that they threaten t...