Natural selection vs genetic drift: two ways populations change

First identify what is changing

An allele is a version of a genetic sequence at a locus. Allele frequency describes the share of copies represented by that version. A genotype describes the combination carried by an individual at that locus. Those are related quantities, but changing how copies are paired can change genotype proportions without changing the allele frequency. Keeping the units clear prevents a large class of confusing evolutionary explanations.

A population-level change also differs from an individual adjusting during its lifetime. The experiments here follow inherited alternatives across generations or construct genotype proportions from allele frequencies. They do not model an organism deliberately producing the variant it needs, or represent every trait as a single gene. Complex traits and real reproductive systems require richer models.

Selection changes expected reproductive contribution

In natural selection, one haploid alternative initially has relative fitness 1.2 and the other one. Fitness is a reproductive contribution within a specified environment. The recurrence weights current frequencies by those contributions. With no finite sampling in this model, the trajectory follows the same deterministic arithmetic every time the starting conditions and environment agree.

Reverse the environment after several generations. The previously favoured alternative now has the lower relative contribution, and its frequency moves in the other direction. This is a useful correction to the idea that evolution is an inevitable ladder of improvement. The model has no universal goal, and its numerical fitness parameter does not express intelligence, moral value or a permanent superiority across environments.

Worked example: change the cause while keeping the start

Both activities begin at fifty-percent frequency, but their update rules answer different questions. The numbers are educational model parameters, not observations from a particular species. Record the rule beside each graph so the visual difference does not become a substitute for the explanation.

1. In natural selection, begin at one half with relative fitness 1.2 versus one. One update gives 0.6 divided by 1.1, approximately 54.5%.
2. In genetic drift, use thirty copies and sample a generation from one half. The realised result must be an integer count divided by thirty, and may move either way.
3. Repeat the drift process, then compare a larger population. Equal fitness remains imposed, so wandering frequencies cannot be attributed to an authored selective advantage.

Drift changes realised frequencies through sampling

The genetic-drift experiment samples the next generation from the previous one. A population of thirty copies does not obtain exactly its expected count every generation. Neutral variants can rise, fall, become fixed or disappear without a fitness difference. Smaller populations tend to experience larger proportional sampling fluctuations under this model; that is a distributional statement rather than a rule for each individual path.

OpenStax's evolution chapter places drift alongside selection, mutation and gene flow. In actual populations these processes can act simultaneously. A founder event can also alter composition without implying the founders were better adapted. Estimating what happened requires evidence beyond noticing that one alternative became common.

Hardy–Weinberg is a reference, not a default explanation

Set the allele frequency to one half in the Hardy–Weinberg experiment. Random union of gametes gives one quarter AA, one half Aa and one quarter aa. Switch to all homozygotes and the proportions become one half AA, zero Aa and one half aa. The allele total remains unchanged. You have altered pairing, not created or removed allele copies.

OpenStax's population-change discussion introduces the equilibrium assumptions. Departures can flag questions worth investigating but do not identify one mechanism on their own. Population structure, nonrandom mating and measurement issues can matter. In finite data, observed proportions also fluctuate around expectations, so exact agreement with textbook fractions is not a reasonable universal requirement.

A mechanism checklist for an evolutionary claim

When you encounter an explanation, ask which causal mechanism is being proposed and what would distinguish it from alternatives. A narrative about adaptation is easier to invent after seeing a result than to test beforehand. Good questions identify evidence that could challenge the story, not just examples that make it sound plausible.

• Distinguish allele frequency, genotype proportion and an individual's phenotype.
• Specify the population, generation interval and inheritance model.
• Ask whether differential reproduction is measured or merely inferred from a rising frequency.
• Consider finite sampling, migration and mutation as alternative or additional mechanisms.
• Separate a mathematical model's expected path from observed biological counts.

Use simple models as contrasts, then restore reality

The models isolate mechanisms to help you recognise what each one can do. That clarity comes from exclusions: the selection board omits drift, while the drift board holds fitness equal and excludes mutation. A real population is not required to choose one of those boxes. The next useful question is how their effects combine, and which measurements let you estimate their relative contribution.

Avoid turning these distinctions into social rankings or just-so explanations of human behaviour. Relative fitness in a toy recurrence is not a general-purpose score of a person or a policy. The valuable habit is more precise: describe the unit, identify the update rule, consider competing mechanisms and state what the evidence leaves unresolved. That habit applies well beyond the biology example without pretending every social process is genetic evolution.