Studying the dynamics of the North American Monsoon System (NAMS) is essential for understanding and assessing the predictability of its variability. Limited-area models are potentially useful tools for this endeavor, but it is important to first identify the suite of physical parameterizations that yields the most realistic simulations. We investigate how different convection and radiation schemes influence simulations of the NAMS produced with the MM5/OSU model. We focus on the simulated intraseasonal variability associated with monsoon onset (June to July) and changes between a wet (1999) and a dry (2000) year. We test six parameterizations, including two convection schema (Grell and Kain-Fritsch) and three radiation schema (CCM2, Cloud, and RRTM). We compare results from 2-month-long simulations to observations of circulation (NCEP Reanalysis) and rainfall (CPC and CMAP). Differences in simulated rainfall produced by the various combinations of schema are substantial, and much greater than the differences that arise from internal model variability in a three-member ensemble of Grell-RRTM simulations. The Grell-RRTM simulation produces the most realistic patterns and magnitudes of rainfall, including intraseasonal variations and the differences between the wet and dry year. Simulations using the Kain-Fritsch scheme produce too much rainfall, and fail to represent the atypical, observed decrease in precipitation from June-to-July in 2000. The CCM2 radiation scheme produces a simulated climate that is too cloudy, yielding little rainfall in the NAMS region regardless of the convection scheme used. The Cloud and RRTM radiation schemes allow for feedbacks between condensation and the water content of clouds, which yields substantial improvements in the model simulations.