Children resemble their parents, yet never exactly. A brown-eyed couple can have a blue-eyed daughter, and a trait can vanish for a generation and then return in a grandchild. People bred animals and crops for thousands of years by choosing the best to mate, but until the late nineteenth century no one could say what actually passed from parent to offspring. Charles Darwin, who built his theory of evolution on inherited variation, admitted the gap plainly in On the Origin of Species (1859): "The laws governing inheritance are quite unknown."

The blending problem

The common assumption in Darwin's day was blending inheritance: offspring are a mixture of their parents, the way red and white paint make pink. Blending seems to fit everyday experience, since many children look roughly halfway between their parents. But it has a fatal flaw. In 1867 the Scottish engineer Fleeming Jenkin pointed out that blending would dilute any new variation by half in each generation, so a rare useful trait would be watered down to nothing long before natural selection could spread it. Darwin took the criticism seriously and never found a satisfying answer. Blending also cannot explain how a trait skips a generation: once paint is mixed, the red cannot come back out.

Mendel's garden

The answer had in fact already been published, in an obscure local journal. Gregor Mendel was an Augustinian friar at the abbey of St Thomas in Brno, in what is now the Czech Republic. Between about 1856 and 1863 he grew and crossed thousands of pea plants in the monastery garden, presenting his results in 1865 and publishing them in 1866. Mendel chose peas wisely. They are easy to grow, they normally fertilize themselves, so pure lines are easy to keep, and they have traits that come in clear either-or forms: round or wrinkled seeds, yellow or green seeds, tall or short stems. Above all, Mendel did something unusual for a naturalist of his time: he counted every kind of offspring and analyzed the ratios.

When he crossed a pure line with yellow seeds with a pure line with green seeds, every plant of the first generation had yellow seeds. Green seemed to have disappeared. But when he let those plants fertilize themselves, green came back in the next generation, in about one plant in four. Across his experiments the ratio was close to three dominant to one recessive. In one cross he counted 6,022 yellow seeds and 2,001 green.

Particles, not paint

Mendel's explanation was that each plant carries two units for each trait, one inherited from each parent, and that these units do not blend. We now call the units genes, and the different versions of a gene are called alleles. One allele can be dominant, showing its effect even when only one copy is present, while a recessive allele shows only when both copies are recessive. A plant with two identical alleles is homozygous; one with two different alleles is heterozygous. The combination of alleles is the genotype, and the visible trait it produces is the phenotype. Two plants can share a phenotype, yellow seeds, while having different genotypes.

Mendel's results suggested two rules, later called his laws. The law of segregation says the two alleles separate when sex cells form, so each sex cell carries only one. The law of independent assortment says alleles for different traits are passed on independently of each other. When Mendel followed two traits at once, seed color and seed shape, the second generation fell into four classes in a ratio of about 9 to 3 to 3 to 1, just what independence predicts. Later research found that the second law has important exceptions, as the next level shows.

Forgotten and found

Mendel's paper was rarely cited for thirty-four years. He became abbot in 1868, was absorbed by administration, and died in 1884. Then in 1900 three botanists, Hugo de Vries, Carl Correns and Erich von Tschermak, independently reported similar results and found that Mendel had got there first. The English biologist William Bateson became Mendel's champion and proposed the word genetics for the new science in 1905. Mendel's data later drew a famous criticism. In 1936 the statistician Ronald Fisher argued that Mendel's ratios fit his predictions more closely than chance should allow: they were too good to be true. Historians still debate whether an assistant tidied the counts, whether Mendel stopped counting when a ratio looked right, or whether Fisher's analysis was too harsh. No one doubts that the laws themselves hold.

Worked example: the test cross

A gardener has a pea plant with yellow seeds and wants to know its genotype. Yellow (Y) is dominant over green (y), so the plant could be YY or Yy. The trick Mendel used is a test cross: breed it with a green plant, which must be yy. If the yellow plant is YY, every offspring receives a Y from it and a y from the green parent, so all are Yy and yellow. If it is Yy, half its pollen carries Y and half carries y, so about half the offspring will be yellow and half green. Suppose the gardener grows forty offspring and eighteen are green. The plant must be Yy; in fact a single green offspring would prove it carries the hidden allele. The same logic, reasoning backward from offspring to the invisible genotype of a parent, is how geneticists read family histories today, because a recessive allele can travel silently through many generations of carriers before two of them meet.