How to Read a Punnett Square

A Punnett square is a bookkeeping device for one assumption: that each parent hands over a randomly chosen copy of each gene, and that the choices are independent of each other. Draw the margins correctly and every box is an equally likely zygote, so the ratio you read out of the boxes is an exact probability rather than an estimate. Draw them wrong — put whole genotypes in the margins instead of gametes, for instance — and the arithmetic quietly goes off by a factor of two. The Punnett square calculator builds the margins from the parental genotypes for you and prints the box count behind every percentage, so you can check the drawing against the number.

What goes in the margins, and what goes in the boxes

Each parent has two copies of a gene and passes on one. So a parent written Aa contributes either A or a, each half the time, and those two possibilities are the parent’s row or column. AA contributes A in both cases — the margin still holds two entries, and that is why the boxes stay equally probable.

Parent	Gametes offered	Probability of each
AA	A, A	1/2 and 1/2
Aa	A, a	1/2 and 1/2
aa	a, a	1/2 and 1/2

The box is the union: one gamete from each parent. With two entries per parent, a single-gene cross always gives four boxes, and each box is worth 1/4. That denominator is the whole reason Mendelian ratios are exact fractions rather than measurements.

The monohybrid square, filled in

Aa × Aa, with the maternal gametes down the side and the paternal across the top:

	A	a
A	AA	Aa
a	Aa	aa

One AA, two Aa, one aa — 25%, 50% and 25%, out of four zygotes. Note that Aa and aA are the same heterozygote reached by two different boxes; that asymmetry is exactly where the 1 : 2 : 1 comes from, and it is the first thing to check when your square refuses to add up to 100%.

The genotype ratio and the phenotype ratio are not the same number

Under complete dominance, one copy of A is enough for the dominant trait, so the two heterozygote boxes join the AA box: 75% dominant, 25% recessive. The classic 3 : 1 is a phenotype ratio sitting on top of a genotype ratio of 1 : 2 : 1, and the two answers are not interchangeable — a genetics problem that asks for either one wants a different line.

Switching the dominance rule moves the phenotype line only, because the boxes never change:

Dominance	Phenotype classes from Aa × Aa	Ratio
Complete	dominant, recessive	3 : 1
Incomplete	dominant, intermediate, recessive	1 : 2 : 1
Codominant	dominant allele shown, both alleles shown, recessive allele shown	1 : 2 : 1

This is why a pedigree that reports “25% affected” and a molecular assay that reports “50% heterozygous” can both be right about the same cross. The Microbiology collection has the same distinction in a different guise: one measurement, two quantities.

Two genes: the 4 × 4 square

With two independently assorting genes each parent offers four gamete types — AB, Ab, aB, ab — so an AaBb × AaBb cross is a sixteen-box square. The box count grows as 4^L: sixteen for two genes, sixty-four for three, two hundred and fifty-six for four, which is why three-gene problems are worked as multiplied single-gene squares rather than as pictures.

Under complete dominance the sixteen boxes collapse to four phenotype classes:

Phenotype class	Boxes	Probability
Dominant at both	9	56.25%
Dominant A, recessive b	3	18.75%
Recessive a, dominant B	3	18.75%
Recessive at both	1	6.25%

The shortcut worth knowing: solve each gene on its own and multiply. 3/4 × 3/4 = 9/16, 3/4 × 1/4 = 3/16, 1/4 × 1/4 = 1/16. Any 9 : 3 : 3 : 1 you derive by the shortcut should match the square box for box; if it does not, a gamete was left out.

The honest caveat is in the assumption. Independent assortment means the two genes are on different chromosomes, or so far apart on the same one that they behave as if they were. Genes that sit close together violate it, and that is a different square — see below.

Sex-linked crosses are read by sex, not by children

An X-linked recessive allele does not fall equally on sons and daughters, because a son gets his single X from his mother and his Y from his father. A carrier mother (XᴺXᶜ) mated to an unaffected father (XᴺY) gives four boxes:

	Xᴺ (father)	Y (father)
Xᴺ (mother)	XᴺXᴺ daughter, unaffected	XᴺY son, unaffected
Xᶜ (mother)	XᴺXᶜ daughter, carrier	XᶜY son, affected

Read within each sex, which is how a counselor quotes it: 50% of sons affected, 0% of daughters affected, 50% of daughters carriers. Read across all four boxes you get 25% — because conceptions are roughly half male and half female, “25% of children” and “50% of sons” are the same fact at two different denominators. Mixing them up is the most common way sex-linked numbers get misreported, and an affected son of this couple did not inherit anything from his father on that chromosome at all.

The conditional question: 2/3, not 1/2

Probabilities change the moment something is already known. An unaffected person with the Aa × Aa parents behind them cannot be aa, so the aa box is discarded and the remaining three renormalize: 1 AA : 2 Aa among the survivors, making the carrier probability 2/3 (66.67%), not the 1/2 someone gets by reading the raw square. The rule generalizes: condition first, then count, then divide by what is left. The calculator’s conditional mode prints the new denominator next to the old one so the shift is visible rather than assumed.

The reverse direction is a Bayesian question and behaves differently. Two healthy children from a couple with a 1% prior carrier chance lower that prior rather than raise it: each unaffected birth multiplies the odds by 3/4, so 1% becomes about 0.565%, and the risk for the next child is roughly 0.14%. One affected child, in contrast, makes both parents’ carrier status certain and returns the flat 25% for the next pregnancy.

Linked genes: when the 9 : 3 : 3 : 1 stops being true

If two loci sit close together on one chromosome, the heterozygous parent produces mostly parental gametes and few recombinant ones. The recombination frequency r is the whole correction — it splits the gametes 1−r into parental types and r into recombinants, and it is what a testcross reads directly, because a homozygous recessive partner contributes nothing that can mask the outcome.

With r = 20% and a coupling-phase heterozygote (AB / ab) crossed to ab / ab:

Offspring class	Gamete type	Probability
AB / ab	parental	40%
ab / ab	parental	40%
Ab / ab	recombinant	10%
aB / ab	recombinant	10%

Two things to take from that table. Recombinants are 20% in total, not 20% each, and 1% recombination is quoted as 1 centiMorgan of map distance — an identity that holds cleanly only for closely linked loci, since double crossovers hide physical distance. Repulsion phase (Ab / aB) flips which classes are the common ones, which is why the arrangement of alleles on the chromosome has to be stated before a ratio means anything.

r = 50% produces exactly the independent-assortment numbers, and that is not a measurement of linkage — it is the absence of one. Anything at or above that value cannot be resolved as linked without far more progeny.

A square is per pregnancy, not per family

The last reading error is the most expensive one. A 25% box is the probability for one child, and a family is a sequence of independent draws. For an Aa × Aa couple planning three children, the chance that at least one is affected is not 75% but 1 − (3/4)³ = 57.8125%; the chance that exactly two are affected is 3 × (1/4)² × (3/4) = 14.06%; and for four children, at least two affected comes to 26.2%. Small families routinely look nothing like the ratio, and that is the arithmetic working correctly, not the model failing.

Before you trust a square

1. Check the margins hold gametes (one allele per gene), not the parents’ genotypes.
2. Count the boxes: 4 for one gene, 16 for two. If your total is not 4^L, a gamete type is missing.
3. Decide which ratio the question asks for — genotype or phenotype — and which dominance rule applies.
4. Confirm the genes are unlinked, or move to a recombination-frequency calculation.
5. For anything sex-linked, state the denominator: within sex, or across all pregnancies.
6. If a phenotype is already known, condition before you divide.
7. If the question is about several children, expand the binomial rather than repeating the single number.

Where the square stops being the tool: allele and genotype frequencies across a whole population are a different question, and knowing which DNA letters you are working with in the laboratory is the practical step before any of it — that is DNA concentration and restriction site territory, in the same Genetics collection.