The science of genetics is concerned with differences between organisms belonging to the same species, or to species which can be crossed. It is not concerned with all such differences. If we have two dogs, one with straight legs, the other with bent legs, the difference may be due to the fact that the bent-legged dog had a diet insufficient in vitamin D, or that his father was a dachshund. The geneticist is concerned with the latter case. The difference may also turn out to be inexplicable in terms of our present knowledge. Now if an observed quantity or quality varies as the result of a number of conditions, we shall try to analyse the variation by keeping all but one of these conditions constant, and varying the remaining one. Thus if we wish to analyse the variation in the volume of a gas we shall first keep its temperature constant, and obtain Boyle’s law; then keep its pressure constant, and obtain Charle’s law. It is obvious that if we are studying the genetics of a character which varies appreciably in response to environment we shall try to keep the environment constant. We shall see that all our dogs receive the same amount of vitamins, sunlight, and so on, so far as possible. This is an ideal which we cannot even approach asymptotically. For it turns out that high frequency radiation and particles of high velocity are very important components of the environment, causing heritable changes by a process called mutation. Even if we could conduct our experiments behind 30 metres of lead the fact that mutation has a temperature coefficient is enough to show that it depends in part on energy fluctuations which are uncontrollable. Secondly we do not know a priori how long our constant environment must operate. In mammals it is no use to stabilise it after birth. For example in the guinea-pig the children of young mothers are more likely than those of old mothers to have extra toes. The question as to how long the environment must remain uniform to eliminate variation due tu [sic] its diversity it [sic] is an empirical question. The affirmation or denial of Lamarckism depends on how it is answered. In a sufficiently constant environment variation still occurs, and it is found that the children of different parents differ. The difference may or may not be an example of heredity in the ordinary sense of that word, namely resemblance between parents and offspring. Thus in Matthiola a yellow flowered plant will give all yellow offspring, and a red plant will give all, or at any rate many, red offspring. Also some parents will give all normal flowered offspring, others a majority with double flowers, in which the presumptive reproductive organs are represented by petals. We cannot however say that doubleness is inherited (I discuss later the meaning of this distressingly vague word), because double-flowered plants are always completely sterile. How are we to minimise the difference between the progeny of different parents? In other words, how are we to approximate to a genetically homogeneous population? The answer to this question is given by experiment, and rests on well-verified theory. There are three main methods. We may obtain a clone, that is to say a group of individuals related only by mitotic, and not meiotic, nuclear divisions. Such are a group of tulips of a named variety derived by vegetative reproduction from a single individual, or a pair of human monozygotic twins. We may obtain a pure line by prolonged inbreeding or by brother sister mating (though here of course we really have a pure line of females and another of males in bisexual organisms). Such are the named varieties of wheat, and a few lines of mice and rats. Finally we may use the first generation of hybrids between two pure lines. It can be shown theoretically that the variance in a population is reduced to a millionth after 10 generations of self-fertilisation or 30 of brother-sister mating. 1 Haldane, JBS. Some Principles of Causal Analysis in Genetics. Erkenntnis 1936; 6: 346-357. Reprinted with the kind permission of Mrs Lois Godfrey.