Sample selection for Cr-rich spinel analysis was performed in several steps. Upon review of all available Leg 38 and all Leg 152 shipboard thin sections, an initial selection of the most promising samples for Cr-rich spinel analysis was made. Every effort was made to sample as closely as possible to intervals of core that had been previously analyzed; sampling was typically done within 10-20 cm of published analyses and always within 50 cm. Splits of these samples were pulverized in a Spex Al-ceramic shatterbox barrel and analyzed by instrumental neutron activation analysis (INAA) to provide a geochemical foundation for placing the Cr-rich spinel analyses in their proper petrological context. Samples (150-200 mg in size) were irradiated for 14 hr in the Texas A&M University (TAMU) TRIGA reactor and counted for 3 hr apiece at 8 to 11 and 40 to 43 days, using facilities in the Center for Chemical Characterization, TAMU. Detailed descriptions of analytical procedures are given in Allan (1995). To ensure run-to-run consistency, a sample of the international basalt standard BHVO was run with each individual group of standards and unknowns, showing excellent overall agreement with accepted values (Table 1) (Gladney and Roelandts, 1988). INAA results for samples chosen for Cr-rich spinel analysis are given in Table 1, complemented by X-ray fluorescence (XRF) analyses by other authors for nearby samples from the same unit. INAA results for samples not chosen for Cr-rich spinel analysis are given in the Appendix.

From this set of initial samples, polished thin sections were made for transmitted and reflective light petrographic analysis. From this set, a smaller subset, covering all Cr-rich spinel-bearing rock types, was selected for major- and minor-element analysis by electron microprobe. The primary criteria for selecting samples was the presence of well-preserved Cr-rich spinels that most likely reflected compositional equilibria with melt upon eruption, rather than spinels that had obviously re-equilibrated with evolving interstitial melt during lava cooling (e.g., coarse-grained samples whose spinels uniformly had thick jackets of chromian magnetite and Ti-rich magnetite were avoided). Electron microprobe analyses were made on the TAMU Geology and Geophysics Department Cameca SX-50 microprobe, equipped with autofocus, an optically-encoded stage, and extensive Sun workstation-based image-processing capabilities. Natural (Jarosewich et al., 1980) and synthetic standards were used, as was a focused beam, sample currents of 10-30 nA, and counting times of 10-180 s. All Cr-rich spinels were initially analyzed by BSE imaging to map out compositional zoning and to avoid regions of obvious Ti-rich magnetite and chromian magnetite during further analysis. Most Cr-rich spinel grains were analyzed by automated rim-to-rim analytical traverses of individual Cr-spinel crystals for all major and minor (>0.2%) elements; very small crystals were checked for zoning by two or three analyses running from core to rim.