The early and middle stages of crystallization of saturated basaltic magma are normally characterized by three reaction series; the two continuous series of plagioclase and of augi te and the discontinuous series olivine- (orthopyroxene-pigeonite). Under high-temperature laboratory conditions there may be complete solid solution throughout the pyroxene field, as indicated by those portions of the field which have been investigated. At the lower temperatures of crystallization which obtain under magmatic conditions, it seems likely that there is limited solid solution in the pyroxene field. Thus two pyroxene phases, a Ca-poor and a Ca-rich phase, will normally crystallize in equilibrium with each other. These two phases may be considered to be diopside-hedenbergite with subordinate clinoenstatite-ferrosilite, and clinoenstatite-ferrosilite with subordinate diopside-hedenbergite in solid solution. The latter phase will appear as ensta-tite-orthoferrosilite if crystallization is below the inversion temperature. During subsequent slow cooling, the pyroxenes tend to assume more ordered structures by exsolution. Thus augites exsolve orthopyroxene or pigeonite and vice versa. Yet, even in the slowly cooled plutonic rocks, complete separation is never attained, though it is approached by exsolution. In volcanic rocks quenching preserves the initial state of solid solution in both phases. Pyroxenes in volcanic, hypabyssal, and plutonic rocks crystallized from saturated basaltic magmas have the same chemical compositions and same degree of solid solution. This is an indication of the rather small range in temperature at which saturated basaltic magmas crystallize, regardless of their environment. Slowly cooled orthopyroxenes exsolve diopside-hedenbergite as fine lamellae parallel to (100). Pigeonites exsolve augite as coarser lamellae parallel to (001). Upon inversion of pigeonite to the orthorhombic form, the augite lamellae remain on the now relict monoclinic (001) plane, and further exsolution of diopsidic pyroxene may take place on (100). Magnesian augites up to approximately Wo40En45Fs15 exsolve ortho-pyroxene as lamellae parallel to (100), but more iron-rich augites exsolve pigeonite parallel to (001). The latter may later invert to hypersthene. These augites may also show a second stage of exsolution of hyper-sthene parallel to (100). In combinations of the slightly dissimilar structures, hypersthene and either augite or pigeonite, the exsolution lamellae tend to be parallel to (100), which is a plane common to both structures. In similar structures, such as augite and pigeonite, the orientation and width of the lamellae are probably controlled chiefly by the ease of access of the ions to the new surface of separation, and hence the preferred orientation is parallel to (001). The authors disagree with Guimaräes' (1948) concept of hypersthenization. In the hypabyssal rocks which he studied, the hypersthenization hypothesis may appear to be valid; but the facts can equally well be explained by the exsolution-inversion hypothesis favored by the writers. Pyroxene relations in plutonic and volcanic rocks can be successfully interpreted on the same basis as hypabyssal rocks by the exsolution-inversion hypothesis, but the hypothesis of hypersthenization cannot account for the pyroxenes in these rocks.