Available evidence indicates that mammary gland secretions first evolved in synapsids that laid parchment-shelled eggs. Unlike the rigid-shelled eggs of birds and some other sauropsids, parchment-shelled eggs lose water very rapidly when exposed to ambient air of lower vapor pressure, whether due to differences in relative humidity or to differences in temperature. This precludes endothermic incubation of parchment-shelled eggs in an open nest. Synapsids may have avoided egg desiccation by incubating eggs in a pouch, but this would limit maternal activity. Parchment-shelled eggs are able to take up liquid water across the eggshell. I propose that mammary secretion originally evolved as a means of supplying water to eggs, and as such was essential to the evolution of endothermy among the egg-laying cynodonts that were ancestral to mammals. It is possible that synapsid eggs, like parchment-shelled squamate eggs, were also capable of uptake of some nutrients, such as sodium and ionic calcium. Living monotremes still produce parchment-shelled eggs. The porous eggshell and bilaminar yolk sac membrane of these eggs permit substantial uptake of uterine secretions during the intrauterine period, and might also facilitate uptake of mammary secretions during egg incubation. In its simplest form, mammary secretion may be an ancient trait of egg-laying synapsids, having had an important role long before milk became obligatory for suckling young.