In populations of finite size, the genetic structure of a new generation is not necessarily that of the previous one. The explanation lies in a sampling effect, based on the fact that a subsample from any large set is not always representative of the larger set. The gametes that form any generation can be thought of as a sample of the alleles from the parental one. By chance the sample might not be random; it could be skewed in either direction. For example, if p = 0.600 and q = 0.400, sampling “error” might result in the gametes having a p value of 0.601 and a q of 0.399. If by chance this skewed sampling occurs in the same direction from generation to generation, the allele frequency can change radically.