
If you start with equal numbers of red and blue beads, will a blind draw give you equal numbers of each? Not necessarily. That gap between the mix you start with and the mix you draw is a way to answer a biology question: What is genetic drift? In a population, the copies of a gene passed on to the next generation need not come in the same proportions as in the previous one. A bead experiment lets you watch those proportions change and compare how much they change in samples of different sizes.
Genetic drift is a change in allele frequency from one generation to the next due to chance. An allele is a version of a gene. Each bead in this model stands for one copy of the same gene: red represents one allele, and blue represents another. If 8 of 10 beads are red, the red allele frequency is 8 ÷ 10, or 80%. The denominator is the total number of gene copies, not the number of individuals.
The starting beads represent the current generation's gene copies. Each draw represents one copy passed on to the next generation. Both colors are equally easy to draw, so neither allele has an advantage. Even so, you can draw more red than blue. As the National Human Genome Research Institute's definition of genetic drift explains, this kind of change comes from chance, not from one allele helping its carriers survive or reproduce.
Use an opaque cup and beads that differ in color but are otherwise alike. To run the small sample, start with 5 red and 5 blue beads. For the large sample, start with 50 red and 50 blue beads. Both start at 50% red. Keep a tally instead of setting each drawn bead aside.
Returning the bead within a round is different from rebuilding the cup between rounds. Within a round, the starting proportions stay fixed. Between rounds, those proportions can change. If you kept drawing from the original 50% mix every time, you could still study sampling variation, but you would not see how one generation's result shapes the next.
Suppose the small sample gives you 7 red draws out of 10, and the large one gives you 57 out of 100. Those are possible results, not predictions. The small sample has moved from 50% to 70% red, a change of 20 percentage points. The large sample has moved from 50% to 57%, a change of 7 percentage points. The large sample actually produced more extra red draws above its expected count: 7 extra rather than 2. But those extra draws make up a smaller share of the total.
To see why, imagine changing just one result from blue to red. In 10 draws, that raises the red frequency by 10 percentage points. In 100 draws, it raises the red frequency by only 1 percentage point. With more draws, there are also more chances for runs of one color to be offset by runs of the other. Large samples can still end up far from the starting frequency, and a small sample can land right on it. The difference is in how widely the results tend to vary over repeated trials.
To test that claim, run several first-generation trials at each size. Reset both cups to 50% red before each trial. Record how far each result is from 50%, regardless of direction. A result of 40% and a result of 60% are both 10 percentage points away. If you averaged the signed changes instead, a big drop could cancel a big rise and hide the variation you wanted to measure. The IBioC guide to interpreting biology graphs can help you compare the spread of the two sets of results.
Now try a different comparison. Start two small cups at 50% red, but this time use each round's tally to build the next generation. One cup might go from 5 red copies to 7 and then back to 5. The other might go from 5 to 3, then to 1, then to 0. Neither path was set by the starting mix. And once a cup has just one red bead, the next 10 draws might miss red completely even though it is still present at the start of the round.
When the red count reaches zero, that allele is lost from the model population. Blue is then fixed, meaning it accounts for every copy of the gene in that population. No amount of mixing can bring back a color that is no longer in the cup. In nature, an allele could reappear through mutation or arrive from another population through gene flow, neither of which this model includes.
If red rises three generations in a row, it is tempting to say red is better. But the bead rules give red no advantage. A rise caused by natural selection would instead involve a heritable difference that affects survival or reproduction. In real populations, selection and drift can both affect the same allele, so seeing its frequency rise does not by itself tell you which process is responsible.
For practice with this kind of reasoning, see the IBioC training and practice problems. Then use your own data: pick two trials that started with the same frequency but ended with different tallies, and explain why neither result requires one color to have an advantage.
Why do I put each bead back after drawing it?
Putting it back keeps the starting mix unchanged during that round. The full tally sets the mix for the next generation.
Can genetic drift happen in a large population?
Yes. Allele frequencies can change by chance in large populations too, but the changes from one generation to the next tend to be smaller.
Can drift make a helpful allele disappear?
Yes. A helpful allele can be lost by chance, especially when only a few copies exist.
About the International Biology Competition: an online biology competition for high school students from around the world. How the rounds work