Navon et al. (1996) demonstrated that the Navon and Stolper (1987) model can be formulated to reproduce a pattern of light-ion lithophile trace element (LIL) enrichments produced by a single, small-scale metasomatic process recorded in a composite xenolith from Dish Hill, California (Nielson et al. 1993). The Navon and Stolper model has failed repeatedly to reproduce the shape and lateral positions of LIL enrichment patterns for samples from peridotite massifs, which are of appropriate scale to test the assumption that LIL fractionation takes place in percolating melts over distances>100 m. The model results also produce unreasonably long times for solidification of thin dikes, which imply untenable thermal conditions for lithospheric mantle.

Using parameters drawn from sample compositions, Nielson et al. (1993) demonstrated, and the calculations of Navon et al. (1996) have shown again, that fractionated trace element patterns of a melt are imprinted upon relatively refractory peridotite matrix in zones closest to a melt source. The observed process sequentially extracts LIL into matrix, analogous to the ion-exchange chromatography of water-purification columns. We have never contended that this process is mathematically distinct from the percolation model of Navon and Stolper (1987), which assumes concentration ofLIL elements in melt. The choice of parameters defines the result, and one would notice a major difference in the taste of water from an ion-exchange column that traps target ions in matrix compared with one that concentrates those ions in the liquid.

The difference between the models is in the selection of parameters and values: The model ofNavon and Stolper (1987) assumes the reaction mechanism, uses theoretical melt compositions, and contains as many as nine unmeasurable parameters. We used the simplified model calculation to avoid reliance on theoretical parameters and to test our assumptions about the process. When the compositions of actual samples are taken as end-members of mantle reactions, the successful results imply that a fractionation-bypercolation process is not applicable to lithospheric mantle. Repetition of the observed small-scale reaction in refractory peridotite must extend the zone of reactions and relative enrichment, centimeter by centimeter, as long as melt aliquots percolate beyond peridotite matrix that had previously reacted to equilibrium with the melt composition. This process satisfactorily explains the wide variations ofLIL fractionation patterns over short distances that characterize mantle rocks in xenoliths and massifs, all of which contain complex systems of mafic intrusions with varied LIL fractionation patterns.