Light scattering and refraction measurements are performed and/or analysed for several polystyrenes and a poly(methyl methacrylate) diluted in various solvents. Contrary to the case of standard polystyrenes in toluene, previously studied, strong deviations occur between the molecular weight values obtained through the classical method M-w(C), through the new proposed method M-w(N), and the expected values M-w (deduced from adequate solutions). In mixtures with large difference between the polymer and solvent refractive indices (for the index increment dn/dc > 0.05 cm(3) g(-1); n, refractive index, c, the concentration): the variation range of M-w(C)/M-w(N) is from about 0.2 to 5. In mixtures with low indices differences, M-w(C)/M-w(N) is systematically strong and even tends to infinity for dn/dc = 0, with M-w(C) tending to infinity, M-w(N) having always reasonable limited values. ;The variation range of M-w(C)/M-w is also infinite (for dn/dc tending to 0), while it remains realistic for M-w(N)/M-w. Three cases are observed for the values obtained by the new method. In a "good" solvent, M-w(N) similar to M-w, in a "bad" solvent, either M-w(N) > M-w or M-w(N) < M-w. The second case can be easily explained by solute aggregates but the third case is more complicated to explain. Eventually, in spite of some necessary progress in understanding the phenomena, the practical prevalence of the new method seems unambiguous. (C) 1999 Elsevier Science B.V. All rights reserved.