This paper attempts to obtain agreement between theoretical and experimental supersonic molecular beam intensities. Measurements of absolute beam intensities are presented, corrected for the effect of background scattering in the expansion chamber. It is found that these intensities are considerably lower than those predicted by existing theories. This is true even when losses due to imperfect skimming are negligible. A new mechanism is therefore proposed which postulates a considerable attenuation of the beam due to "self-scattering" collisions upstream of the skimmer and downstream of the onset of translational freezing. Using this mechanism it is possible to reasonably reproduce the experimental intensities.