Here is the evolution of an allele in which the starting frequency is 0.2 and the selective advantage is 0.01. The population remains constant at 5000 throughout the 400 generations. The final allele frequency distribution is wide, with a broad peak centered around 0.4.
Below is the corresponding case when the population hits a "bottleneck". Population size is suddenly reduced from 5000 to 100 at generation 160. (Think dinosaurs.) Everything else is the same as in the model above. The variance increases at this time and by generation 400 most of the realizations result in loss of the allele in the population. However, in almost all the cases in which the allele is not lost, it becomes fixed in the population. The 'control' model (large, unchanged population, above) would eventually (~1000 generations or more) end up with the allele becoming fixed in the population, but the bottleneck vastly speeded up the evolution.
Finally, I looked at a case in which an allele has negative selection advantage, which would eventually die out in a population with a continuously large population. The variance caused by the bottleneck at generation 160, when the population drops from 5000 to 100, results in a significant number of realizations (about 25%) in which the allele becomes fixed in the population, quite the opposite of its fate had there not been a bottleneck. How cool is that!
