The compositions of natural glasses and phenocrysts in basalts from Deep Sea Drilling Project Sites 501, 504, and 505, near the Costa Rica Rift, constitute evidence for the existence of a periodically replenished axial magma chamber that repeatedly erupted lavas of remarkably uniform composition. Magma compositions were affected by three general components: (1) injected magmas carrying (in decreasing order of abundance) Plagioclase, olivine, and chrome-spinel phenocrysts (spinel assemblage); (2) injected magmas carrying Plagioclase, clinopyroxene, and olivine phenocrysts, but no spinel (clinopyroxene assemblage); and (3) moderately evolved hybrids in the magma chamber itself. The compositions of the injected phenocrysts and minerals in glomerocrysts are as follows: Plagioclase-Angj,^; olivine-Fo 87 _ 89 ; clinopyroxene-high Cr 2 O 3 (0.7-1.1%), endiopside (Wc^E^jFs^, and aluminous chromian spinel (Cr/Cr + Al 0.3). These minerals resemble those thought to occur in upper mantle sources (9 kbars and less) of ocean-ridge basalts and to crystallize in magmas near those sources. In the magma chamber, more sodic Plagioclase (An 79 _g 5 ), less magnesian olivine (Fo 8J _ 86 ) and low-Cr 2 O 3 (0.1-0.4%) clinopyroxene formed rims on these crystals, grew as other phenocrysts, and formed cumulus segregations on the walls and floors of the magma chamber. In the spinel-assemblage magmas, magnesiochromite (Cr/Cr + Al = 0.4-0.5) also formed. Some cumulus segregations were later entrained in lavas as xenoliths.The glass compositions define 16 internally homogeneous eruptive units, 13 of which are in stratigraphic order in a single hole, Hole 504B, which was drilled 561.5 meters into the ocean crust. These units are defined as differing from each other by more than analytical uncertainty in one or more oxides. However, many of the glass groups in Hole 504B show virtually no differences in TiO 2 contents, Mg/Mg + Fe 2 " 1 ", or normative An/An + Ab, all of which are sensitive indicators of crystallization differentiation. The differences are so small that they are only apparent in the glass compositions; they are almost completely obscured in whole-rock samples by the presence of phenocrysts and the effects of alteration. Moreover, several of the glass units at different depths in Hole 504B are compositionally identical, with all oxides falling within the range of analytical uncertainty, with only small variations in the rest of the suite. The repetition of identical chemical types requires (1) very regular injection of magmas into the magma chamber, (2) extreme similarity of injected magmas, and (3) displacement of very nearly the same proportion of the magmas in the chamber at each injection. Numerical modeling and thermal considerations have led some workers to propose the existence of such conditions at certain types of spreading centers, but the lava and glass compositions at Hole 504B represent the first direct evidence revealed by drilling of the existence of a compositionally nearly steady-state magma chamber, and...