Punnett Squares Explained: Monohybrid Crosses, Genotypes, and Probability

A Punnett square is a grid that lists every combination of the alleles two parents can pass on. Each cell is one possible offspring genotype, and counting the cells gives the genotype and phenotype ratios for the cross. It is the simplest model in genetics, and the most misunderstood: its percentages describe many offspring, not a promise about one family. This guide covers monohybrid and dihybrid crosses, how to read the ratios, and the limits of a model that assumes one gene per trait.

How a Punnett square works: gametes, cells, counting

Each parent contributes one allele per gene to each offspring, through a gamete. A parent with genotype Aa makes two kinds of gamete in equal numbers: A and a. The square lists one parent’s gametes along the top and the other’s down the side, and each cell combines the row and column to give one possible offspring genotype. For Aa × Aa the grid is 2 × 2: AA, Aa, Aa, aa. Counting gives the genotype ratio 1:2:1. If A is dominant, the first three cells show the dominant trait and the last shows the recessive one, giving the phenotype ratio 3:1.

The letters are arbitrary labels, and which allele is called dominant is a choice about the trait, not about the letters. Swapping the labels — calling the recessive allele A instead — does not change the genotype counts; it only renames which cells are called dominant. That is why the calculator asks which allele is dominant: the genotype ratio is fixed by the parents, and the phenotype ratio follows from the dominance choice. A square where changing the dominant allele changed the genotype counts would be broken.

Homozygous parents make the square trivial, which is itself informative. AA × aa gives four Aa cells: every offspring is a heterozygote, and every one shows the dominant trait. Aa × aa gives 2 Aa : 2 aa, a 1:1 ratio. The square’s value is not in hard cases but in making the counting visible: every outcome, once, in its cell, with nothing hidden in a formula.

The dihybrid square: sixteen cells, nine genotypes, four phenotypes

Two genes mean each parent makes four kinds of gamete. A parent who is AaBb makes AB, Ab, aB, and ab gametes in equal numbers, assuming the genes assort independently — they sit on different chromosomes, or far apart on the same one. The square is 4 × 4, sixteen cells. Counting genotypes gives nine kinds: 1 AABB, 2 AABb, 1 AAbb, 2 AaBB, 4 AaBb, 2 Aabb, 1 aaBB, 2 aaBb, 1 aabb. Counting phenotypes, with both A and B dominant, gives 9 showing both dominant traits, 3 showing only the first, 3 showing only the second, and 1 showing neither: the classic 9:3:3:1 ratio.

The dihybrid ratio is really two monohybrid ratios multiplied together, which is why it factorises so neatly. The chance of the dominant phenotype for gene A is 3/4 and for gene B is 3/4, so both together is 3/4 × 3/4 = 9/16. That multiplication is the product rule of probability, and it is the same rule the Punnett Square Calculator applies when it reports phenotype shares for multi-trait crosses. Three traits would need a 64-cell grid, which is why the calculator, and this guide, stop at two: the counting scales, but the grid stops being readable.

Independent assortment is an assumption, and it fails when genes are linked — close together on the same chromosome, so they tend to be inherited together. Linked genes give ratios skewed toward the parental combinations, and the skew measures the distance between the genes. The square still works as a starting point, but its equal-probability cells are wrong for linked genes. Real traits are also usually polygenic, governed by many genes plus environment, which the next section takes up.

Probability, not promise: what the ratios do and do not mean

A 25% chance means that, over very many offspring, about a quarter will show the outcome. It does not mean the fourth child will, or that one child in every four must. Each child is an independent draw from the same grid. The chance that a 25% outcome never appears in four children is (3/4)⁴ = 31.6% — nearly a third of four-child families. A small family that looks nothing like the ratio has not broken the rules; it has sampled them, and small samples deviate.

The same logic applies in reverse. A family of four children all showing a 75% trait is unremarkable: (3/4)⁴ = 31.6% again. The ratio describes the population of possible children, not the destiny of actual ones. This is the single most important idea in the whole topic, and it is the one the square’s neat percentages most often obscure. Read the ratio as a long-run frequency, and the temptation to predict one child from it disappears.

Sex-linked traits, codominance, and incomplete dominance all bend the simple picture without breaking the counting. In codominance both alleles show, as in AB blood group; in incomplete dominance the heterozygote is intermediate, as in pink snapdragons from red and white parents. The grid still lists the genotypes correctly — the change is in how genotypes map to phenotypes. The calculator assumes simple dominance for one gene per trait, and says so. Anything more complex needs a model that names its own assumptions.

Trait examples, and the limits of one gene per trait

Eye colour and hair type are the textbook examples, and both are warnings. Eye colour is influenced by several genes, with brown dominant in the simple model but hazel, green, and blue arising from combinations the one-gene square cannot produce. Hair type — straight, wavy, curly — is similarly polygenic, and the square’s prediction for a single gene is at best a rough sketch. The kept questions in the evidence ask how to do a Punnett square for eye colour; the honest answer is that the square teaches the method, and the real trait needs more genes than the square holds.

The model also assumes no mutation, no selection, no linkage, and equal survival of all genotypes. For classroom crosses of pea plants, those assumptions hold well enough that Mendel’s ratios appear. For human families, they hold approximately, and the square is a teaching and counselling aid, not a prediction. It does not predict what a child will look like, and any use that claims otherwise — “what will my baby look like” calculators that output a face from a square — is astrology with a grid. The square predicts ratios over many offspring, under stated assumptions, and nothing more.

Used within its limits, the square is exact. Its cells are counts divided by grid sizes, and they are exact within the model. When a real family deviates, the deviation is sampling, not rounding, and the correct response is to remember the sample size, not to question the decimals. A square that is read with its assumptions in view is one of the clearest tools in biology; read without them, it is one of the most over-claimed.

Linked genes: when the square’s assumption breaks

The dihybrid square assumes the two genes assort independently — that the allele a gamete carries for one gene says nothing about the allele it carries for the other. That assumption holds when the genes sit on different chromosomes, or far apart on the same chromosome. When two genes sit close together on the same chromosome, they are linked: they tend to travel together into the same gamete, and the offspring ratios deviate from 9:3:3:1 toward the parental combinations. The square’s arithmetic is still correct for the gametes that are formed; what changes is the gamete frequencies, which are no longer equal.

Recombination is the exception that measures the linkage. During meiosis, chromosomes exchange segments — crossing over — and a crossover between two linked genes swaps their alleles, producing recombinant gametes. The closer the genes, the rarer the crossover between them, and the rarer the recombinants. Genetic distance is measured in centimorgans, where 1 centimorgan corresponds to a 1% recombinant frequency, and a linkage map is built from exactly these frequencies. Two genes 10 centimorgans apart produce about 10% recombinant gametes and 90% parental ones, and the offspring ratios follow those frequencies, not the equal quarters of the independent square.

The practical consequence is that a Punnett square for linked genes must be drawn with the actual gamete frequencies, not the equal ones. List the four gamete types with their frequencies — parental types at (1 − r)/2 each and recombinant types at r/2 each, where r is the recombination frequency — and build the square from those. The cells are then weighted by the gamete frequencies, and the phenotype ratios follow. For genes on different chromosomes, r = 0.5, the frequencies are equal, and the familiar 9:3:3:1 returns. The square is a framework; the gamete frequencies are the input, and linkage is the case where the input must be measured rather than assumed.

Frequently asked questions