The purpose of this work is to examine physical radiation dose differences between two multileaf collimator (MLC) leaf widths (5 and 10 mm) in the treatment of CNS and head and neck neoplasms with intensity modulated radiation therapy (IMRT). Three clinical patients with CNS tumors were planned with two different MLC leaf sizes, 5 and 10 mm, representing Varian-120 and Varian-80 Millennium multileaf collimators, respectively. Two sets of IMRT treatment plans were developed. The goal of the first set was radiation dose conformality in three dimensions. The goal for the second set was organ avoidance of a nearby critical structure while maintaining adequate coverage of the target volume. Treatment planning utilized the CadPlan/Helios system (Varian Medical Systems, Milpitas CA) for dynamic MLC treatment delivery. All beam parameters and optimization (cost function) parameters were identical for the 5 and 10 mm plans. For all cases the number of beams, gantry positions, and table positions were taken from clinically treated three-dimensional conformal radiotherapy plans. Conformality was measured by the ratio of the planning isodose volume to the target volume. Organ avoidance was measured by the volume of the critical structure receiving greater than 90% of the prescription dose (V90). For three patients with squamous cell carcinoma of the head and neck (T2-T4 NO-N2c MO) 5 and 10 mm leaf widths were compared for parotid preservation utilizing nine coplanar equally spaced beams delivering a simultaneous integrated boost. Because modest differences in physical dose to the parotid were detected, a NTCP model based upon the clinical parameters of Eisbruch et al. was then used for comparisons. The conformality improved in all three CNS cases for the 5 mm plans compared to the 10 mm plans. For the organ avoidance plans, V90 also improved in two of the three cases when the 5 mm leaf width was utilized for IMRT treatment delivery. In the third case, both the 5 and 10 mm plans were able to spare the critical structure with none of the structure receiving more than 90% of the prescription dose, but in the moderate dose range, less dose was delivered to the critical structure with the 5 mm plan. For the head and neck cases both the 5 and 10 × 2.5 mm beamlets dMLC sliding window techniques spared the contralateral parotid gland while maintaining target volume coverage. The mean parotid dose was modestly lower with the smaller beamlet size (21.04 Gy vs 22.36 Gy). The resulting average NTCP values were 13.72% for 10 mm dMLC and 8.24% for 5 mm dMLC. In conclusion, five mm leaf width results in an improvement in physical dose distribution over 10 mm leaf width that may be clinically relevant in some cases. These differences may be most pronounced for single fraction radiosurgery or in cases where the tolerance of the sensitive organ is less than or close to the target volume prescription. © 2002 American Association of Physicists in Medicine.