Genetic drift is random change in allele frequencies caused by finite sampling from one generation to the next.

Even when two alleles have identical effects on fitness, not every individual contributes the same number of descendants by chance. The next generation therefore receives a sample of the current gene pool rather than an exact copy of its frequencies.

Simple sampling example

Suppose allele $A$ has frequency $p=0.50$ in a diploid population of only 10 individuals, so there are 20 allele copies. If the next generation also contains 20 copies, random reproduction need not produce exactly 10 $A$ and 10 $a$ copies. It might produce 13 $A$ and 7 $a$, giving

$$p' = \frac{13}{20}=0.65.$$

Nothing in this example makes $A$ advantageous. The frequency changed because of sampling chance.

Drift is stronger in smaller populations

Sampling fluctuations are proportionally larger when fewer allele copies are passed on. Large populations tend to show smaller random changes per generation, whereas small populations can change rapidly.

Repeated drift can eventually lead to:

• fixation, when an allele reaches frequency 1;
• loss, when it reaches frequency 0;
• reduced genetic variation within a population.

Drift and natural selection can operate simultaneously. Selection is systematic with respect to fitness differences; drift is stochastic sampling noise. When selection is weak and populations are small, drift can overwhelm the directional effect of selection for substantial periods.

Genetic drift therefore provides a crucial counterexample to the idea that every evolutionary change is an adaptation.