The variation of the probability of a son within and across couples

It is suggested that there is a flaw in the currently accepted account of the variation of P, the probability of a boy, within and across couples. It was previously suggested that P has a mean (for Caucasian couples) of approximately 0.514 with an SD of approximately 0.05 across couples: and that the variation within couples is rather less. Grounds are offered here for suspecting that this formulation underestimates both SDs by a factor of as much as 4. It is suggested that in estimating these sources of variation, earlier workers did not consider the possibility that within-couple variation might be random and substantial. In view of the established epidemiology of human sex ratios, it now seems likely that such variation exists, and that there is a substantial measure of counterbalancing across-couple variation.