How can genetic drift change a small population by chance?

Genetic drift is a change in how common different versions of a gene—called alleles—are in a population, caused by chance. Think of each generation as a new sample: only some parental alleles end up in offspring. That sampling can be uneven even when no allele is better than another. Drift is especially important in small populations, where a few chance outcomes can make a larger difference to the overall proportions. [S1][S2]

So “by chance” does not mean that the population has no pattern at all. It means the frequency shift is not driven by an allele’s advantage in reproductive success.

Suppose, as a hypothetical example, a small population has 10 diploid individuals. At one gene, imagine there are 20 total allele copies: 10 copies of allele A and 10 of allele a. Each allele starts at a frequency of 10/20, or 0.50.

Now suppose that, by chance, the next generation has 13 A copies and 7 a copies. The frequency of A is now 13/20, or 0.65. A rose from 0.50 to 0.65, but this example gives no reason to think A helped its carriers. It could simply be a chance difference in which alleles were passed on. The numbers illustrate the idea; they are not a prediction about what any real population will do. [S2]

The key difference is what drives the change. With drift, allele frequencies shift through chance. With natural selection, frequencies shift because traits are associated with differences in reproductive success. [S1]

For instance, if A becomes more common in the example, drift alone does not show that A is useful. To argue for selection, you would need evidence connecting a trait associated with A to differences in reproductive success. Drift and selection can also occur in the same population at the same time, so a frequency change alone does not establish which process caused it. [S1]

Keep the explanation focused on the population, the allele frequencies, and the proposed cause. A useful copyable example is:

“Assume a population starts with [number] copies of allele A out of [total copies], so A’s frequency is [starting frequency]. In the next generation, there are [new number] A copies out of [new total], so its frequency is [new frequency]. If the change is due to chance sampling rather than a reproductive advantage, it illustrates genetic drift. The frequency change by itself does not prove selection or establish the cause in a real population.”

For a real population, a before-and-after comparison by itself may not tell you why frequencies changed. The sources describe drift and selection as distinct processes that can occur together; they do not provide a method for determining the cause in a particular population. [S1]