We are presenting here an updated version of our predictive model of solvation parameters of solutes, based on
a simplified molecular topology (J. Chem. Inf. Model, 2006, 46, 1723-1734). The learning experimental database of this
model is presently larger than in our first version and includes more compounds with more than one functional group of
the same type. This experimental database is also surer, in the sense that it only includes compounds in liquid state at
room temperature, when the polarizability of a compound in solution is established via its refractive index in gas or crystalline
state. Indeed, we demonstrate in this paper, a bias in this case.