Experimental melt (glass) compositions indicate that most of the S-type granites of the Cape Granite Suite in South Africa have ferromagnesian contents too high to represent melts. Consequently, the composition of the more mafic granites demands the addition of an Fe- and Mg-rich component to the magma. The compositions of the granites evolve along well-defined trends away from the likely melt composition for many components plotted against Mg + Fe. An increase in A/CNK, Mg#, Ca, and HREEs, as well as a decrease in K and Si, as a function of increasing Mg + Fe appears to limit the contaminant to garnet (up to 20 wt%). The rate of Ti increase, as a function of Mg + Fe increase in the granites, matches that defined by the stoichiometry of high-temperature biotite, but cannot be the product of accumulation of biotite (± other phases) in the magma because the chemical trends are inconsistent with this, particularly those portrayed by K and Ti as a function of Mg + Fe. This, in conjunction with the fact that no large, counterbalancing population of very leucocratic material exists in the Cape Granite Suite, suggests that the relatively mafic granites are not the products of garnet fractional crystallization. Rather, these appear to be the result of selective entrainment of peritectic garnet and ilmenite. Thus, this work indicates that much of the compositional variation in the granites is primary, reflecting the magma composition that ascended from the source, and is controlled by the proportion of peritectic products entrained into the melt. There is no indication of entrainment of a mineralogically diverse residuum (restite).