In a recent experimental study of rare earth element (REE) partitioning between immiscible silicate liquids, Ellison and Hess (1989) applied the two-lattice melt component activity model of Nielsen (1985) in an attempt to resolve the compositional dependence of REE partitioning. On the basis of their results they stated that the model failed to remove the compositional dependence. Upon examining their results we find that the application of the Nielsen (1985) two-lattice melt model removes 90-96% of the compositional dependence. Interpretations of melt structure based on application of the two-lattice model are nebulous due to the simplifying assumption of ideal intra-lattice mixing. Nevertheless, a comparison of the results of several versions of the model indicate that in this system, Ti behaves more like a network former than a network modifier. In addition, the assumption of Ca-Al complexing does not improve the effectiveness of the model at resolving compositional dependence over the assumption that excess Al acts as a network modifier.