We review our methodology for producing physically accurate potential energy functions, particularly relevant in the context of Lifson's goal of including frequency agreement as one of the criteria of a self-consistent force field. Our spectroscopically determined force field (SDFF) procedure guarantees such agreement by imposing it as an initial constraint on parameter optimization, and accomplishes this by an analytical transformation of ab initio "data" into the energy function format. After describing the elements of the SDFF protocol, we indicate its implementation to date and then discuss recent advances in our representation of the force field, in particular those required to produce an SDFF for the peptide group.