Résumé : We present a comparative analysis of the PISA and RAM models using Monte Carlo (MC) simulations as references. The PISA and RAM models have in common a factorization of the pair distribution into an angular and a radial part. For the angular part both the PISA and RAM models use a Boltzmann weighting. First, we show that this factorization itself and the Boltzmann weighting are satisfactory approximations in a wide range of density and temperature conditions. Indeed, the Boltzmann weighting yields very good average energies (angular average). The chief approximation of the PISA method consists of taking a Heaviside step function for the distribution of the centres of mass. In the RAM theory the distribution of the centres of mass in the molecular fluid is supposed to be the same as the distribution of a spherically symmetric reference fluid. The RAM hypothesis and the PISA approximation are tested and discussed for several molecular liquids (nitrogen, bromine, carbon tetrachloride, benzene, and cis- and trans-decalin). We show that the RAM theory is unable to give a reliable radial distribution function except for systems studied under favourable conditions (nitrogen and bromine at high temperature). The agreement between MC and RAM radial distribution functions is particularly poor for highly non-spherical molecules like trans-decalin. The quality of the PISA approximation of the radial distribution function is also discussed in terms of molecular shape. Finally, we show that the more sophisticated RAM theory does not necessarily yield better internal energies than the simplest but very efficient PISA model. © 1990 Taylor & Francis Group, LLC.