Monte Carlo evaluations and the source normalisation problem.
It is well known that Transport Theory calculations give results which are immediately intelligible, as they take into account the source explicitly. Monte Carlo evaluations on the contrary are given per source particle, and they have to be normalized according to the actual source to be completely intelligible. Unfortunately in MC evaluations sometimes one can be led, to reach good results from a statistic point of view (that is low variance values), to manipulate somehow the source instead of sharply define it, in order either to fully exploit the features of his own code or to bypass and simplify actual physical difficulties, but on the other hand this procedure strongly increases the complexity of the source normalization process, which becomes (or can become) the true problem indeed. Two examples will be analyzed: a shielding problem of a transportation cask which will be investigated by means of MCNP, and a shielding problem of a proton beam which will be investigated by means of FLUKA. In both the problems a correct, sharply defined description of the source leads to hardly manageable situations, when "tricky" adjustments on the source instead lead smoothly to statistically good results while the source normalizations are anyway solved without too much complexity.