Spinels in four Apollo 12 igneous rocks have been analyzed with the electron microprobe. Spinel composition is complex and reflects solid solution between five end members that can best be represented in a modified spinel prism in which Ti4+ replaces Fe3+. Some rocks (12036, 12038) contain two main spinel groups with divergent compositional trends. In addition, 12036 contains rare hercynite-rich spinels in pyroxene melt inclusions; these have not previously been reported in lunar samples. Rock 12039 contains only one ulvöspinel-rich late spinel group. Rock 12051 has spinels which are often continuously zoned from chromite- to ulvöspinel-rich end members with no compositional discontinuities. Conclusions: (1)A three-dimensional representation of the complex spinel solid solutions in Apollo 12 rocks and a knowledge of textural and paragenetic relationships of spinels and coexisting phases are necessary for interpreting their petrogenetic significance. (2) The presence of two groups of spinels with divergent trends, commonly found in Apollo 12 igneous rocks, is the result of crystallization of early chromite-rich spinel and olivine, terminated by a peritectic reaction of the type described by Irvine (1967), and later crystallization of ulvöspinel-rich spinel, either simultaneously with, or later than, pyroxene. The “gap” in composition between the two spinel groups is the result of a change in composition of the melt during the course of crystallization. In rock 12039, complete resorption of early spinel (and olivine) may have resulted in the formation of only one, late spinel group. In 12051, rapid crystallization of early chromite-rich spinel simultaneously with pyroxene, rather than olivine, results in continuously zoned crystals, trending towards the Fe2TiO4 end member. No divergence in compositional trends is noted, apparently because there is no intervening peritectic reaction. (3) No evidence is found in support of several other hypotheses suggested in explanation of the divergent Apollo 12 spinel trends, such as the presence of an immiscibility field, or reaction of early chromite-rich spinel and melt to yield late ulvöspinel-rich spinel. (4) Melt of the composition of rock 12038 could not be the parent of an olivine-pyroxene cumulate rock such as 12036 (or 12040) because early 12038 spinels differ markedly from early 12036 spinels. However, a melt intermediate in composition between 12036 and 12038 could be parental to both rocks. The residual melt of this parent, after olivine and pyroxene crystal settling, would be enriched in feldspar components and this could explain the unusual major and trace element composition of rock 12038.