Two gene versions, called alleles, start out at exactly 50% each in a small population. Nothing favours either one. Nobody is fitter, healthier or luckier in any way that matters. Yet if you wait long enough, one of them will vanish and the other will take over completely. No selection, no pressure, just chance. This is *genetic drift*, and you can watch it happen below.
Think of the next generation as a handful of marbles drawn from a bag. The bag holds the gene copies of the current generation: half blue, half orange. Draw 20 marbles and you will rarely get exactly 10 of each. You might get 12 blue, or 8, or 14. Nothing is wrong with the bag. A small sample just does not mirror it perfectly.
Real populations work the same way. Each parent passes on only some of its gene copies, and which ones get through is down to chance. So the share of each allele wobbles from one generation to the next. In a big population the wobble is tiny. In a small one it is large.
The wobble has no memory and no pull back towards 50%. If blue drifts up to 60%, the next draw starts from 60%. And if blue ever reaches 100%, orange is gone for good, because a draw cannot produce a marble that is no longer in the bag. Both 0% and 100% are one-way doors.
Each line below is one population. Every generation, the whole population is redrawn at random from the one before. Lines that hit the top have lost the orange allele (lime). Lines that hit the bottom have lost the blue one (orange). Try shrinking the population, then try making it huge.
Each individual carries two gene copies, so a population of N has 2N copies in the bag. Changing a slider restarts the run.
With 10 or 20 individuals, most lines hit a wall within a few dozen generations. With 500, the lines barely leave the middle in the time you have. Same rules, same chance, very different outcome. The only thing you changed was the size of the sample.
First, it eats variety. In the simple textbook model, the share of individuals carrying two different copies of a gene falls by about 1 in 2N each generation. Halve the population and variety disappears twice as fast.
Second, the odds of winning are fair. Set the starting share of blue to 20% and run it a few times. Blue takes over in roughly one population in five, and the rule behind that is simple: for a gene with no advantage, the chance it eventually takes over equals its starting share. A brand-new mutation in a population of N is one copy out of 2N, so its chance is 1 in 2N. Most new neutral mutations are lost within a few generations. A rare few win by luck alone.
In the 1950s the American geneticist Peter Buri tested this with fruit flies. He used two versions of an eye-colour gene, *bw* and *bw75*, which make flies with different eye colours. He set up about a hundred separate bottles, each starting at 50% of each allele. In every bottle, every generation, he kept just 16 flies, 8 males and 8 females, and let them breed. He followed the bottles for 19 generations.
All the bottles started the same, and the environment was the same. Yet the bottles drifted apart. Some ended up with only one allele, either one, and others sat at all sorts of mixtures in between. It looks just like the chart above, only with real flies.
Drift in a real population shows up when the population gets squeezed. Around 1775 a typhoon hit Pingelap, a tiny atoll in the Pacific, and left only about 20 survivors. The island's ruler at the time is believed to have carried the gene for achromatopsia, a condition that causes complete colour blindness. With so few people left, his copy of the gene had a large share of the new population, and the population then grew from that small starting point.
Today almost 10% of Pingelapese have achromatopsia, and around another 30% are carriers who do not show it. The gene gained no benefit from the typhoon. It simply made it through a narrow gap, and the narrower the gap, the more chance matters. Biologists call this a *bottleneck*, or the *founder effect* when a few settlers start a new population.
Drift is not the same as natural selection. Selection pushes an allele up or down because it helps or harms survival. Drift pushes it in random directions regardless. In large populations selection usually wins out, because chance averages out. In small ones, drift can overpower even a mildly harmful allele and carry it to fixation, or lose a mildly helpful one. Both forces are at work all the time. What decides who is winning is, once again, the size of the sample.
Next time you hear that a gene is common, remember that it may be common because it was useful, or because it got lucky.