The sensitivity of a mesoscale model to different microphysical parameterizations is investigated for two events of precipitation in the Mediterranean region, that is, the Mesoscale Alpine Program (MAP) intensive observation periods (IOP) 2b (19–21 September 1999) and 8 (20–22 October 1999). Simulations are performed with the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5); the most commonly used bulk microphysical parameterization schemes are evaluated, with a particular focus on their impact on the forecast of rainfall. To evaluate the forecast skill, the verification is carried out quantitatively by using the observations recorded by a high-resolution rain gauge network during the MAP campaign. The results show that, for the surface rainfall forecast, all microphysical schemes produce a similar precipitation field and none of them perform significantly better than the others. The ability of different schemes to reproduce events with different ongoing microphysical processes is briefly discussed by comparing model simulations and knowledge of hydrometeor fields from radar observations. The vertical profiles of hydrometeors from two of the analyzed schemes show gross similarities with available radar observations. Last, the role of one of the parameterizations appearing in a typical bulk microphysical scheme, that is, the one of the snowfall speed, is evaluated in detail. Adjustments in the semiempirical relationships describing the fall speed of snow particles have a large impact, because a reduced snowfall speed enhances precipitation on the lee side of mountain ridges and diminishes it on the windward side. Anyway, this effect does not appear to be able to largely improve or reduce the forecast skill of the MM5 systematically; the impact of changes in the parameterization of the snow deposition velocity very likely depends on the dynamics of the event under investigation.