Major-element analyses of several thousand glasses from all of the Apollo landing sites have resulted in the identification of 25 groups of pristine (i.e., volcanic) glass. The nickel in these pristine glasses is shown to be indigenous, not meteoritic contamination, and to be correlated with Mg. The chemical data indicate that these glasses are consistently better candidates for primary magmas than the majority of crystalline mare basalts. The pristine glasses support the view that assimilative processes [Ringwood and Kesson, 1976] involving two cumulate systems in the differentiated mantle operated during mare petrogenesis. The reality of those assimilative interactions is evident by the occurrence of two linear arrays among the chemistries of the glasses. Data suggest that these cumulate components in the differentiated mantle persist for lateral distances of at least 1000 km and therefore appear to be products of a vast magma ocean that existed early in lunar history. Alternative hypotheses claiming that the highlands crust and mare source regions were produced by serial magmatism without a magma ocean have not yet considered all of the constraints derived from the pristine glasses.