Population bottlenecks are often linked to explain low levels of genetic variation in natural populations, yet few studies have documented the direct genetic consequences of known bottlenecks in the wild. Empirical studies of natural population bottlenecks are therefore needed, because key assumptions of theoretical and laboratory studies of bottlenecks may not hold in the wild. Here we present microsatellite data fr om a severe bottleneck (95% mortality) in an insular population of song sparrows (Melospiza melodia). The major findings of our study are as follows: (i) The bottleneck reduced heterozygosity and allelic diversity nearly to neutral expectations, despite non-random survival of birds with respect to inbreeding and wing length. (ii) All measures of genetic diversity regained pre-bottleneck levels within two to three years of the crash. This rapid recovery was due to low levels of immigration. (iii) The rapid recovery occurred despite a coincident, strong increase in average inbreeding. These results show that immigration at levels that are hard to measure in most field studies can lead to qualitatively very different and genetic outcomes from those expected from mutations only. We suggest that future thee empirical work on bottlenecks and metapopulations should address the impact of immigration.