The development of the ocean crust is one of the most fundamental processes operating within the earth. In the study of these processes geochemical data are important in constraining the primary igneous behaviour, alteration, and in determining the nature of chemical fluxes. With the low core recoveries typically encountered by the Ocean Drilling Program (ODP) in basement drilling log data are of particular importance in providing essentially continuous geochemical data through the drilled section. This contribution presents an interpretation of the geochemical and petrological evolution of the basaltic crust in Hole 896A which was drilled 200 km south of the Costa Rica Rift. The recovered lithologies (pillow lavas, sheet flows, breccias) have all suffered alteration and veining to a greater or lesser degree. The quantitative measurement of geochemistry downhole relies on the application of nuclear logging techniques. In the ODP Schlumberger's Geochemical Logging Tool (GLT (super *) ) has often been deployed, and this measures the major elements: Si, Al, Ti, Fe, Ca, K and S. Subsequent processing converts the initial element yields into oxides by use of a closure model. In this model the assumption is made that the sum of the measured oxides (i.e. SiO (sub 2) +Al (sub 2) O (sub 3) +TiO (sub 2) +FeO+CaO+K (sub 2) O) is equal to the average recovered basalt (XRF whole rock) minus the (unmeasured) MgO+Na (sub 2) O value. Unfortunately this average basalt is inevitably derived from "fresh" material, which is not representative of all the drilled material. GLT data from Hole 896A has a large compositional range, with end-members unexpected form oceanic basement. Part of this compositional spectrum reflects poor data quality in regions of borehole break-outs but the remainder is the result of processing artefacts and in particular is related to the closure model applied. In reality the range of MgO+Na (sub 2) O ranges from near zero (carbonates) to over 20% (saponite veins) as compared to the average 10% assumed in the processing. In this contribution an attempt is made to remove this affect from the long data through the use of assumed mineralogy which is then all closed to the same value. From this approach it is possible to demonstrate that the cloud of GLT data has a series of definite vectors which relate to both primary and secondary processes.