What a Punnett square is really doing
A Punnett square is a probability table. Reginald Punnett drew it in 1905 as a bookkeeping device: list every gamete one parent can make down the side, every gamete the other can make across the top, and each cell is one equally likely fertilisation event. Count the cells that give you the genotype you care about, divide by the total, and you have its probability.
The square works because of two of Mendel's laws. Segregation says the two alleles at a locus separate during meiosis, so a heterozygous Aa parent makes gametes that are half A and half a. Independent assortment says what happens at one locus tells you nothing about another, so an AaBb parent makes four gamete types — AB, Ab, aB, ab — in equal quarters. Every cell in the resulting 4 × 4 grid is therefore equally likely, which is the only reason counting cells gives you a probability at all.
That equal-likelihood property is also the square's limitation. The moment two loci sit close together on the same chromosome, the four gamete types stop being equally frequent and the square gives the wrong answer. That case needs a recombination frequency, handled by the recombination frequency and map distance calculator.
Why you should multiply instead of drawing the grid
Beyond one locus, drawing the square is the slow way. A trihybrid cross needs a 64-cell grid; a four-locus cross needs 256. The product rule gets the same answer in a line of arithmetic, because independent assortment makes the loci independent events.
Handle one locus at a time. In Aa × Aa, the offspring are 1/4 AA, 1/2 Aa, 1/4 aa — so 3/4 show the dominant phenotype. In Aa × aa, they are 1/2 Aa and 1/2 aa. Then multiply across loci. The chance that an offspring of AaBb × AaBb shows both dominant traits is 3/4 × 3/4 = 9/16, and the chance it is aabb is 1/4 × 1/4 = 1/16. Those are the 9 and the 1 of the famous 9:3:3:1.
The same logic gives you the structural numbers without any counting. A parent heterozygous at n loci makes 2n gamete types; the square has 4n cells; a self-cross of two such parents yields 3n genotypes and, under complete dominance, 2n phenotypes. The probability of the fully dominant phenotype is (3/4)n and of the fully recessive one (1/4)n.
Dominance changes the phenotype column, never the genotype column. Under complete dominance AA and Aa are one phenotype, so 1:2:1 collapses to 3:1. Under incomplete dominance or codominance the heterozygote is distinguishable, so the phenotype ratio stays 1:2:1 — which is why a pink snapdragon self-cross gives red, pink and white in a 1:2:1 ratio rather than 3 red to 1 white.
Worked example: AaBb × AaBb, seed shape and seed colour
Cross two dihybrid pea plants, heterozygous for round seeds (A dominant to wrinkled a) and yellow cotyledons (B dominant to green b). Score 16 seeds.
- List the gametes. Each parent is heterozygous at two loci, so each makes 22 = 4 gamete types:
AB,Ab,aB,ab, each with probability 1/4. - Size the square. 4 × 4 = 16 cells, each with probability 1/16.
- Work locus A alone.
Aa×Aa→ 1/4AA, 1/2Aa, 1/4aa. Dominant phenotype = 1/4 + 1/2 = 3/4. - Work locus B alone. Identical: dominant phenotype = 3/4, recessive = 1/4.
- Multiply for round and yellow. 3/4 × 3/4 = 9/16 = 56.25%.
- Round and green. 3/4 × 1/4 = 3/16 = 18.75%.
- Wrinkled and yellow. 1/4 × 3/4 = 3/16 = 18.75%.
- Wrinkled and green. 1/4 × 1/4 = 1/16 = 6.25%.
- Check. 9/16 + 3/16 + 3/16 + 1/16 = 16/16 = 1. The phenotype ratio is 9:3:3:1.
- Expected counts in 16 seeds. 9, 3, 3 and 1 respectively — the reason a 16-seed sample is the classroom default.
Now the genotypes. Locus A gives three classes in 1:2:1, and so does locus B, so the cross yields 3 × 3 = 9 genotypes in the ratio 1:2:1:2:4:2:1:2:1, summing to 16. The chance of an offspring being heterozygous at both loci — genotype AaBb, the most common single genotype — is 1/2 × 1/2 = 4/16 = 25%. Notice that the most common genotype is not the most common phenotype: nine of the sixteen seeds are round and yellow, but only four of them are AaBb.
How to read the ratios, and what a deviation means
Match the ratio to the cross that produces it. A 3:1 phenotype ratio in the progeny means both parents were heterozygous at that locus. A 1:1 ratio means one parent was heterozygous and the other homozygous recessive — that is a test cross, and it is the standard way to determine whether a dominant-looking individual is AA or Aa. A uniform F1 with no segregation at all means at least one parent was homozygous. A 1:2:1 phenotype ratio means the heterozygote is visibly distinct, so dominance is incomplete or the alleles are codominant.
Expect the observed counts to miss. Sixteen seeds will rarely come out exactly 9:3:3:1; that is sampling, not a failed hypothesis. The question is whether the deviation is larger than sampling alone would plausibly produce, and the standard test for it is a chi-square goodness-of-fit against the expected ratio. Run yours through the genetics chi-square calculator before concluding that a gene is misbehaving.
Ratios that are not powers of the standard ones signal real biology. A 9:7 or 9:3:4 or 12:3:1 phenotype ratio from a dihybrid self-cross is still built from the same 16 cells — the classes have merged because the two genes interact. That is epistasis, and the merged pattern tells you how. A 2:1 ratio in a monohybrid cross usually means the homozygous dominant class is lethal, so a quarter of the offspring are missing before you count.
Square size and probabilities by number of heterozygous loci
| Loci (n) | Gametes per parent | Cells in the square | Genotype classes | Phenotype classes | P(all dominant) | P(all recessive) |
|---|---|---|---|---|---|---|
| 1 (monohybrid) | 2 | 4 | 3 | 2 | 75% | 25% |
| 2 (dihybrid) | 4 | 16 | 9 | 4 | 56.25% | 6.25% |
| 3 (trihybrid) | 8 | 64 | 27 | 8 | 42.1875% | 1.5625% |
| 4 | 16 | 256 | 81 | 16 | 31.6406% | 0.3906% |
| 5 | 32 | 1,024 | 243 | 32 | 23.7305% | 0.0977% |
Phenotype classes assume complete dominance at every locus. Under incomplete dominance the phenotype classes equal the genotype classes, 3ⁿ.
Where Punnett square answers go wrong
- Assuming the loci assort independently when they are linked. Genes close together on one chromosome co-segregate, so the parental gamete types are over-represented and the square's equal-cell assumption fails.
- Reading a genotype ratio as a phenotype ratio. Under complete dominance the 1:2:1 genotype ratio and the 3:1 phenotype ratio describe the same 4 cells. State which one you mean.
- Expecting the observed counts to match exactly. A litter of four from a 3:1 cross produces all four dominant offspring about 32% of the time, purely by chance.
- Forgetting sex linkage. An X-linked locus needs a square with hemizygous male gametes; sons and daughters have different expected ratios and must be counted separately.
- Treating a probability as a guarantee. A 25% risk applies independently to each child. Three unaffected children do not make the fourth more likely to be affected.
- Using single-gene logic on a quantitative trait. Height, yield and skin colour are governed by many loci plus environment, and no Punnett square describes them.
- Ignoring penetrance and expressivity. A genotype that does not always produce its phenotype breaks the mapping between the genotype ratio and what you can actually see.
When to reach for a different tool
The Punnett square answers a question about one family: given these two parents, what can this cross produce? Population questions need different machinery. If you want to know how common a genotype is across a whole population rather than in one mating, the Hardy-Weinberg equation is the right model — start with the Hardy-Weinberg equilibrium calculator, or count genotypes directly with the allele frequency calculator.
Clinical risk questions usually start from a probability that a parent is a carrier rather than from a known genotype, and they combine that prior with test results. That is a Bayesian calculation layered on top of the Punnett square's 1/4, and it belongs in the autosomal recessive carrier risk calculator.
If all you need is the number of distinct gametes a genotype can produce — the row and column count before you draw anything — the gamete combinations calculator gives it directly. And once you have real progeny counts in hand, testing them against the ratio predicted here is a separate step with its own statistics.
Key terms
- Genotype
- The pair of alleles an individual carries at a locus, written AA, Aa or aa. Two individuals with different genotypes can share a phenotype.
- Phenotype
- The observable trait. Under complete dominance the phenotype does not distinguish AA from Aa, which is what collapses a 1:2:1 genotype ratio into 3:1.
- Test cross
- A cross to a homozygous recessive individual, used to reveal whether a dominant-looking individual is homozygous or heterozygous. AA × aa gives uniform offspring; Aa × aa gives 1:1.
- Independent assortment
- Mendel's second law: alleles at different loci segregate into gametes independently. It holds for loci on different chromosomes and fails for tightly linked loci.
- Incomplete dominance
- Inheritance in which the heterozygote shows an intermediate phenotype, so the phenotype ratio equals the genotype ratio.
- Epistasis
- Interaction in which one gene masks or modifies the effect of another, merging Punnett square classes and producing ratios such as 9:7, 12:3:1 or 9:3:4.