The principal thrust of this work is to explore the validity of the major assumptions made in developing the solvent triangle and to examine the relationship between the selectivity parameters derived from the solvent triangle and independently determined, solvatochromically based measures of dipolarity, hydrogen bond acidity, and basicity. It is shown that the original solvent triangle classification scheme is only slightly modified by the use of a much more rigorous correction for the behavior of a reference alkane. These results are based on new experimental measurements of the gas-liquid partition coefficients of Rohrschneiders' probes solutes and a set of alkanes (n-pentane-n-octane). More fundamentally we have shown that the original selectivity parameters based on the properties of ethanol, dioxane and nitromathane are all lumped parameters composed of dipolar, hydrogen bond acidity and hydrogen bond basicity terms. Perhaps most importantly the original probe solutes used to develop the solvent traingle are shown to be inefficient choices in terms of their ability to discriminate between similar solvents. This is an important limitation in that the primary use of the solvent triangle has been for the optimization of selectivity and the classification of phases.