A conspicuous deficiency of practically all birth order studies is the absence of a general population control group. Instead, a number of statistical methods of estimating the expected number of subjects in each sibling position have been used. These are based upon the assumption that for any given size of sibship there should be equal numbers of subjects in each birth rank. As Price and Hare (1969) have pointed out, there are a number of reasons why such as assumption is incorrect, and therefore why the conclusions based upon it are unreliable. Because, in most studies, only relatively small numbers of patients are available various methods of combining sibships of different sizes are adopted and the sexes of the siblings are ignored. To consider a sibling position irrespective of the size and composition of the sibship in which it occurs is unjustifiable. As birth order effects are relatively slight, large numbers of subjects are required to detect them. Unless a patient sample of several thousand is used the numbers in any sibship size are too small to permit analysis by birth rank. As sibship size is itself a possible variable, and as within each sibship the sexes of the siblings can be distributed in a variety of ways, birth order studies should ideally consider one sibship size at a time and should investigate sibling position by sex of sibling. Unfortunately, in sibships of four or more the numbers of combinations become so great as to render this approach impracticable.