Genetic drift

Genetic drift is the change in allele frequency in a population due to random sampling — the statistical noise inherent in reproducing a finite number of individuals from a finite number of parents. It is not a force of selection, not a bias toward fitness, but the consequence of the fact that populations are not infinite and reproduction is not deterministic.

This is not an error term to be ignored in evolutionary models. It is a central evolutionary mechanism, and in many populations — especially small ones — it is the dominant one.

The Measurement Problem

Genetic drift was not predicted by theory and then confirmed by observation. It was forced on evolutionary biology by recalcitrant data. Early population geneticists expected allele frequencies to stabilize at values determined by selection coefficients. Instead, they fluctuated. Populations of Drosophila in controlled laboratory environments, with constant selection pressures, still showed variation in allele frequencies across replicates. The environment was held fixed; the genes were not.

Sewall Wright interpreted this as evidence that random sampling matters. R.A. Fisher did not. The dispute was not over mathematics — both agreed on the binomial sampling formula — but over whether the effect was large enough to dominate real evolutionary dynamics. Wright said yes in small or subdivided populations. Fisher said no in large, panmictic ones. The data vindicated Wright, but it took decades and the arrival of molecular evidence to settle it.

Effective Population Size

The strength of drift is inversely proportional to effective population size (N_e), not census population size. A species with a million individuals but extreme reproductive variance — where most offspring come from a tiny fraction of adults — experiences drift as if the population were far smaller. N_e is what matters, and N_e is almost always smaller than the headcount suggests, sometimes by orders of magnitude.

This has consequences. Alleles with small selective advantages (s < 1/2N_e) behave as if neutral — drift dominates their dynamics. In a population of N_e = 1,000, an allele conferring a 0.01% fitness advantage is effectively invisible to selection. It will drift. Most populations are not large enough for most mutations to be resolved by selection.

Neutral Theory and the Molecular Clock

In the 1960s, molecular biologists began sequencing proteins. They expected to find that most amino acid differences between species were adaptive. Instead, they found that most substitutions occurred at a roughly constant rate — the molecular clock. Motoo Kimura proposed that most observed substitutions at the molecular level are neutral or nearly neutral, fixed by drift rather than selection. The rate of substitution is then determined not by adaptive advantage but by mutation rate and genetic drift.

This was not a claim that most mutations are neutral in effect (most are deleterious), but that most substitutions — mutations that go to fixation — are neutral. Selection filters out the bad; drift fixes the invisible. The result is a molecular evolutionary process dominated not by adaptation but by stochastic sampling.

The neutral theory remains controversial in its strong form, but its core insight is empirically robust: a large fraction of observed molecular evolution is not explainable by selection. Drift is not a footnote. It is the null hypothesis.

Founder Effects and Bottlenecks

When a population is founded by a small number of individuals — a founder event — or crashes to a small size and recovers — a bottleneck — drift becomes extreme. Allele frequencies in the new population are a random sample of the old one, and rare alleles are often lost. The result is reduced genetic diversity and the fixation of alleles that may have been rare or neutral in the ancestral population.

Humans went through at least one severe bottleneck roughly 70,000 years ago, possibly associated with the Toba supervolcano eruption. The genetic signature is unmistakable: low diversity compared to other great apes, consistent with descent from a small founding population. We are a drifted species.

Interaction with Selection

Drift does not replace selection. It competes with it. In large populations, selection dominates; in small ones, drift does. The boundary is determined by the product N_es: when this is much larger than 1, selection wins; when much smaller, drift wins. Most real populations sit in the intermediate regime where both matter.

This has a perverse consequence: traits that are slightly deleterious can fix by drift in small populations, even in the face of selection against them. The result is not adaptation but genetic load — an evolutionary burden imposed by the statistical structure of reproduction. Natural selection does not always optimize. Sometimes it loses to noise.

Provocation

The traditional narrative of evolution is a narrative of adaptation: organisms evolving solutions to environmental problems, features honed by selection. Genetic drift is treated as a qualifier, a minor complication in an otherwise adaptationist story. The empirical record suggests the opposite. Drift is not the exception; it is the null case. Most alleles are born neutral, live neutral, and die neutral, their fates determined by the stochastic arithmetic of sampling. Selection is the intervention, the rare event that pulls a lineage away from the random walk.

If you believe that most of what you see in biology is the product of natural selection, you are not reasoning from evidence. You are reasoning from intuition about design. The data say otherwise.