# Drift, Gene Flow and Competing Explanations

> Across replicate populations, drift spreads allele frequencies apart without pushing their average; gene flow and consistent selection push the average. This lesson builds that rule and uses it to tell chance from cause. It treats genetic drift as sampling error, follows it through Buri's 107 fruit-fly lines, island founders and bacterial colonies, and then shows gene flow both eroding and rescuing populations. Afterwards you can compute how migration shifts a frequency and how much drift scatters it, and design the replicates, neutral null and fitness assay that separate drift from selection.

Course: Ecology and Evolution (Evidence and evolutionary reasoning). Level: core university (third year). About 20 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/drift-gene-flow-explanations/

Builds on: How to Test an Evolutionary Explanation · Natural Selection in Ecological Context

## What you'll learn

- Define genetic drift as chance change in allele frequency from random sampling of parents, and explain founder effects, bottlenecks and effective population size (Ne) as distinct from census size.
- Explain how gene flow can swamp local adaptation (migration load in Timema walking-sticks) or supply useful variation (genetic rescue of Trinidadian guppies; adaptive introgression in Gulf killifish).
- Calculate the deterministic shift caused by migration and the standard deviation of the allele frequency produced by one generation of neutral sampling.
- Given one pattern, state the drift prediction and the selection prediction and the evidence that separates them (replicates, a neutral null, an independent fitness assay).

The first film built the rule with bags of beads; this text adds its limits, the size of the scatter, and real populations. The key tool is **replicate populations**: populations started at the same allele frequency and kept alike, so chance shows up as a spread among them.

> **The one idea.** Across replicate populations, drift spreads allele frequencies apart without pushing their average; gene flow and consistent selection push the average. So to tell the forces apart, compare the replicates’ average with a **neutral null** (what chance alone predicts), then test selection with an independent fitness assay and gene flow with measures of immigration and ancestry.
>
> *Scope:* one **locus** (a position or region in the genome), two alleles, replicates that start alike in one setting. “Without pushing” means the expected average stays put; a finite set of replicates still wobbles, with no trend. Gene flow moves each population a fraction m of its distance from the immigrants’ frequency pm, so with the same m and pm, populations that start alike move alike and populations that have drifted apart do not. Consistent selection favours the same allele in every replicate. One population’s frequency, on its own, cannot tell luck from cause, and an unchanged average does not exclude selection (Going deeper). This scope holds wherever the rule appears below.

## Drift is sampling error

The first film’s bead draw stands for two chance steps in a real population: which of a parent’s two copies goes into each egg or sperm, and how many offspring each parent leaves. The resulting change in allele frequency is **genetic drift**. It needs no difference in **fitness** (the relative number of surviving offspring an individual leaves in a given environment), so it acts on all alleles alike and can carry a harmful allele up or a beneficial one to loss, especially in small populations.

The textbook model of a closed, ideal population is the **[Wright-Fisher](https://stemhop.com/methods/population-genetic-models/) model** (assumptions in Going deeper): N diploid individuals carry **2N gene copies**, and each generation’s 2N copies are drawn at random from the parents’ pool. A bigger sample is a closer copy of the pool, so the scatter shrinks as 2N grows:

> **Equation card: If the same generation could be rerun many times: where its allele frequency would land on average, and how widely the reruns would scatter around that.**
>
>     expected next p = p variance of next p = p(1 − p) / 2N
>
> In words: On average the frequency stays where it was; the variance is p times one minus p, divided by the number of gene copies sampled.
>
> - p: the allele's frequency among the parents
> - 1 − p: the other allele's frequency
> - 2N: gene copies sampled: two per diploid individual (copies)
>
> Sanity check: As 2N grows without limit the scatter goes to zero. At p = 0 or 1 it is also zero: a fixed allele cannot drift.

**Predict.** A neutral allele sits at 0.60. If drift runs until only one allele is left, how often will this one be the survivor?

> **Check your prediction.** 60% of the time. Far enough in the future, every copy of the gene descends from a single copy alive now. Under neutrality that ancestor is a random pick, and 60% of today’s copies carry this allele. So a neutral allele’s chance of **fixing**, reaching frequency 1, equals its current frequency; at 0.60 it is lost 40% of the time. [Population genetics](https://stemhop.com/lessons/population-genetics/) in the Genetics course works through the algebra.

Real populations drift faster than their **census size** N, the head count, suggests: if a few parents leave most of the offspring, the next generation samples only those few, and an uneven breeding sex ratio or a size that swings from year to year does the same. So drift is scaled by the **effective population size (Ne)**, the size of the ideal population that drifts as fast, usually below N.

A **bottleneck** is a sharp drop in numbers that leaves a chance sample of the original gene pool. A **founder effect** occurs when a new population starts from a few individuals who carry an unrepresentative sample of their source population’s variation.

**Worked problem** (illustrative numbers). A neutral allele has frequency 0.60 in a resident population and 0.20 among immigrants. After migration, immigrants make up m = 0.10 of the breeding pool. The next generation is 100 diploid offspring, so 2N = 200 gene copies are sampled, with no selection, no mutation and no further migration during sampling.

1. *Migration.* The mixed frequency is a weighted average: p = 0.90 x 0.60 + 0.10 x 0.20 = 0.54 + 0.02 = 0.56. Gene flow shifted the frequency by 0.56 - 0.60 = -0.04, a deterministic shift, the same every time.
2. *Drift’s expected outcome.* 200 x 0.56 = 112 copies, a frequency of 0.56; drift adds no push.
3. *Scatter from drift.* The variance is 0.56 x 0.44 / 200 = 0.2464 / 200 = 0.001232. Variance is in squared units, so take its square root, the **standard deviation**, to get the scatter in frequency units: 0.035, about 3.5 percentage points.

**Predict.** Cut the offspring tenfold, to 10 (2N = 20 copies). Is the scatter ten times bigger?

> **Check your prediction.** No, about three times (the square root of 10 is 3.16). The variance, 0.2464 / 20 = 0.01232, is ten times bigger, but the standard deviation is its square root: 0.111, about 11 points.

*Interactive on the lesson page:* One generation: migration, then sampling

One generation in one population gives a single draw from that scatter. To see the scatter itself you need many populations at once, which is what Buri built.

**Film transcript: 107 Bottles**

### Film transcript: 107 Bottles

To see drift's scatter, you need many populations at once. Do real animals spread apart like the beads, with their average left alone? From one bottle's frequency alone, you could not tell. Follow one line alone. It climbs, dips and climbs again. Was it pushed, or was it luck? Peter Buri published the real experiment, with fruit flies. He followed two alleles of a gene for eye colour, the brown locus. Every line started at one half gold. Each generation, he picked sixteen flies as the parents of the next. His first series had a hundred and seven lines. Every bottle was set up the same way and kept the same way. So chance alone predicts differences between the lines.

Now stack all the lines into one picture. Each bar counts how many lines sit at each frequency. At the start, every line is in the middle bar. Generation by generation, the pile flattens. Lines drift up and down. Some lose blue entirely, some lose gold, and those stay at the edges. By generation ten, the edges are filling up. Now watch the white triangle, the pooled frequency across all lines. By generation nineteen, will it be up, down, or still at one half?

Still at about one half. By then many lines had ended with only one allele. Yet the pooled frequency showed no systematic change. Spread without a push on the average. That is the pattern drift predicts, now seen in real flies rather than beads. But how fast did the lines spread?

Buri asked how big an ideal population would have to be to spread that fast. Here is the spread computed for an ideal population of sixteen. Buri's real lines spread faster than that, with more of them at the edges. From how fast his own lines spread, Buri estimated about nine. They drifted as if each bottle held only nine flies. That number is the effective population size, Faster spread means a smaller effective size. It measures how hard chance acted, not how many flies were in the bottle. So one line wanders, and tells you little about cause. Many replicates, side by side, show whether the average was pushed.

From the film

- Buri started 107 fruit-fly lines at an allele frequency of 0.5 and kept 16 parents in each, generation after generation.
- Over 19 generations the lines spread toward 0 and 1, and many ended with only one allele.
- The frequency pooled over all the lines showed no systematic change: spread without a push on the average, the pattern drift predicts.
- From how fast the lines spread, Buri estimated an effective size of about 9, below the 16 flies in each bottle.
- One line on its own wanders and tells you little about cause; replicates show whether the average was pushed.

## Buri’s 107 fly lines: drift as a distribution

Two things the second film left out: how firm its reading is, and why nine.

The pattern is consistent with drift rather than proof of it; Buri also compared the changes with what sampling alone predicts and found no evidence of selection. His Ne of 9.0, estimated from how fast the lines spread apart, is the mean for his first series. Buri said the cause of the gap was not known, but judged that variation in how many offspring each fly left, down to complete sterility, was certainly part of it, and thought crowding for egg-laying space in the small vials probably played a role.

## Founders and frontiers: anoles on islands, bacteria on agar

Founder effects in nature are usually reconstructed after the fact; Jason Kolbe, Jonathan Losos and colleagues made them happen, releasing one randomly drawn male-female pair of brown anoles, Anolis sagrei, from a single source island onto each of seven small islands in the Bahamas, where the vegetation was narrower than on the source.

**Predict.** One source, random founders, narrow perches everywhere: will the island populations come to look alike, stay different, or both?

> **Check your prediction.** Both. The 2012 report in Science found that genetic and morphological differences that tracked the founding pairs persisted throughout the experiment: that is the spread. And all the populations evolved shorter hindlimbs, the direction predicted for lizards on narrow perches: a shared shift in the average, the pattern selection predicts.

*Figure (AI-generated illustration).* Seven small Bahamian islands with narrow vegetation, each founded by one random brown anole pair from a single source: founder effects show as islands that differ, selection as a shift they all share. Schematic, not to scale. Image: a larger source island on the left with broad shrubs and brown anole lizards, with gold arrows to seven small islands on the right, each with slender vegetation and one male and one female lizard. An inset compares a broad source branch with a thin twig on the small islands.

Large populations drift too. Oskar Hallatschek and colleagues (2007, PNAS) mixed two Escherichia coli strains that differed only in a fluorescent label and let the colony spread across agar. It broke into single-colour sectors, because chance in the thin band of pioneer cells at the edge decided which colour got carried outward: a spreading front is a series of founder events. As with Buri’s flies, the head count is not what chance acts on.

*Figure (AI-generated illustration).* Only a thin band of pioneer cells at the edge founds each new stretch of colony, so chance in that band, not the mass of cells behind it, decides which colour is carried outward, even in an enormous population. Schematic. Image: a round colony on an agar plate whose speckled two-colour centre gives way to sharp single-colour wedges radiating outward, with a thin gold band marking the growing edge. Two insets show the mixed start and a zoom on the few pioneer cells at the rim.

## Gene flow: swamping or rescue

**Gene flow** is the movement of alleles between populations by individuals or gametes, pollen included. Beyond the first film’s matched bags, each population moves by m × (pm − p), a fraction m of its distance from the immigrants’ frequency pm (step 1 of the worked problem). With m = 0.10 and pm = 0.20, a population at 0.30 moves to 0.29 and one at 0.70 to 0.65: populations that had drifted apart move by different amounts, and the gap shrinks from 0.40 to 0.36. Two-way exchange also pulls populations together.

It can harm. **Migration load** is a drop in a population’s mean fitness caused by immigrant alleles that suit somewhere else. Timema cristinae walking-sticks live on two host plants, and colour morphs differ in how well they are camouflaged on each. Daniel Bolnick and Patrik Nosil (2007, Evolution) compared 25 natural populations and found that those receiving more immigrants from the other host were more poorly matched, and that these showed larger within-generation shifts in morph frequency from selection. It is a correlation that fits the migration-load prediction, not a manipulation of migration.

**Predict.** Two small, isolated guppy populations in Trinidad start receiving immigrants. Will their fitness fall, as the walking-sticks’ did?

> **Check your prediction.** No, it rose. In the Caigual and Taylor streams, gene flow began after fish from a different source were introduced upstream. Sarah Fitzpatrick and colleagues (2020, Current Biology) followed the recipients with pedigrees. Both populations grew nearly tenfold, and on average hybrids lived longer and reproduced more than residents or immigrants, though in some comparisons hybrids and pure immigrants did about equally well.

That makes this **genetic rescue**: a rise in a small population’s fitness or growth caused by the genetic effects of gene flow, not just by extra bodies (demographic rescue); the fit hybrids in the pedigrees point to a genetic cause. Nor did the immigrants simply replace the residents: alleles putatively associated with local adaptation were not entirely swamped.

Help can also come from another species: **[introgression](https://stemhop.com/methods/introgression-tests/)** is the movement of alleles from one species into another through hybridization followed by backcrossing, hybrids mating back into the receiving species. In the Houston Ship Channel and Galveston Bay region of Texas, Gulf killifish in polluted water carry tolerance-associated variants that Oziolor and colleagues (2019, Science) inferred had entered by introgression from Atlantic killifish: gene flow supplied the variation, and strong selection raised it.

Whether gene flow harms or helps depends on what the incoming alleles do in their new home.

## Three explanations for one pattern

Back to the island allele. Drift spreads replicates without pushing their average; gene flow and consistent selection push it. Each leaves its own signature:

| Explanation | Replicate populations in the same setting | Fitness assay |
| --- | --- | --- |
| Drift alone (the neutral null) | Scatter both ways; the average stays near the starting frequency | Carriers do no better (the model assumes equal fitness) |
| Gene flow | Frequencies track the amount and source of immigration | Not the test: measure immigration and ancestry |
| Consistent selection | The average shifts toward the favoured allele; drift still scatters each one | The favoured genotypes do better (for a recessive allele, only its homozygotes) |

The **neutral null** can be [Wright-Fisher](https://stemhop.com/methods/population-genetic-models/) or a colony model, and the **independent fitness assay** must not rely on the pattern itself. Often the answer is drift and selection together, as with the anoles. The lessons on [testing an evolutionary explanation](https://stemhop.com/lessons/testing-evolutionary-explanations/) and [selection in ecological context](https://stemhop.com/lessons/selection-in-ecological-context/) use the same logic.

### Design problem

In expanding bacterial colonies, one fluorescent strain forms large sectors. Is it fitter, or did it get lucky at the frontier? Design the experiment.

> **One way to do it.** Grow many replicate colonies from one starting mix, say 20 at 50:50. Under the neutral expectation Weinstein and colleagues (2017, PLOS Computational Biology) used, large sectors in a few colonies are what drift produces; a consistent shift in the mean fraction, beyond what a neutral model calibrated to your colonies allows, is evidence of selection.
>
> Then remove two confounds: swap the fluorescent labels and repeat, so any cost of carrying a marker shows up; and measure each strain’s expansion speed growing alone, an independent fitness assay that does not depend on the sectors.
>
> Keep the scope: Weinstein and colleagues found that some strains ranked differently on agar than in well-mixed liquid culture. A strain that wins on a plate has been shown to win on a plate.

## Going deeper

### The Wright-Fisher assumptions

A closed population of N diploid individuals, constant in size, with non-overlapping generations, random mating, no selection, no mutation and no migration. Real populations break these, which is why Ne differs from N.

### Drift erodes variation

In the neutral [Wright-Fisher](https://stemhop.com/methods/population-genetic-models/) model with no mutation or migration, expected **heterozygosity** (the chance that two gene copies drawn at random differ; under random mating, the expected share of heterozygous individuals) falls each generation, faster when Ne is small, though one population’s can wobble. Mutation and balancing selection can maintain it. So small, isolated populations lose variation, which gene flow can restore.

### When does selection beat drift?

A rough rule: when Ne multiplied by the selection coefficient s (an allele’s proportional fitness advantage or cost) is well below 1, drift dominates; when it is well above 1, selection does. Do not read 1 as a sharp crossover: in the standard additive model, a new beneficial allele with Ne × s = 1 is already about four times as likely to fix as a neutral one. With few replicates, weak selection can hide inside the average’s chance wobble; more replicates, or more generations sampled through time, raise the power to detect it but do not guarantee it.

### Selection that leaves the average alone

Balancing selection, such as a heterozygote advantage whose balance point is the starting frequency, can hold the replicate average still while keeping the spread narrower than drift alone would.

### When there are no replicates

For a single island or [introgression](https://stemhop.com/methods/introgression-tests/) event, a time series, a genome-wide neutral null, migration estimates and fitness assays can still test whether chance alone explains the pattern.

### Where Buri’s design came from

Buri, publishing in 1956, ran a large version of a laboratory design that Warwick Kerr and Sewall Wright had introduced in 1954, using Drosophila melanogaster and two alleles at the brown eye-colour locus.

### Founders and new species

Whether founder effects alone can produce new species is debated, and the anole experiment did not test that.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/drift-gene-flow-explanations/
