Respiratory motion can introduce substantial dose errors during IMRT delivery. These errors are difficult to predict because of the nonsynchronous interplay between radiation beams and tissues. The present study investigates the impact of dose fractionation on respiratory motion induced dosimetric errors during IMRT delivery and their radiobiological implications by using measured 3D dose. We focused on IMRT delivery with dynamic multileaf collimation (DMLC-IMRT). IMRT plans using several beam arrangements were optimized for and delivered to a polystyrene phantom containing a simulated target and critical organs. The phantom was set in linear sinusoidal motion at a frequency of 15 cyclesmin (0.25 Hz). The amplitude of the motion was ±0.75 cm in the longitudinal direction and ±0.25 cm in the lateral direction. Absolute doses were measured with a 0.125 cc ionization chamber while dose distributions were measured with transverse films spaced 6 mm apart. Measurements were performed for varying number of fractions with motion, with respiratory-gated motion, and without motion. A tumor control probability (TCP) model for an inhomogeneously irradiated tumor was used to calculate and compare TCPs for the measurements and the treatment plans. Equivalent uniform doses (EUD) were also computed. For individual fields, point measurements using an ionization chamber showed substantial dose deviations (-11.7% to 47.8%) for the moving phantom as compared to the stationary phantom. However, much smaller deviations (-1.7% to 3.5%) were observed for the composite dose of all fields. The dose distributions and DVHs of stationary and gated deliveries were in good agreement with those of treatment plans, while those of the nongated moving phantom showed substantial differences. Compared to the stationaryphantom, the largest differences observed for the minimum and maximum target doses were -18.8% and +19.7%, respectively. Due to their random nature, these dose errors tended to average out over fractionated treatments. The results of five-fraction measurements showed significantly improved agreement between the moving and stationary phantom. The changes in TCP were less than 4.3% for a single fraction, and less than 2.3% for two or more fractions. Variation of average EUD per fraction was small (<3.1 cGy for a fraction size of 200 cGy), even when the DVHs were noticeably different from that of the stationary tumor. In conclusion, IMRT treatment of sites affected by respiratory motion can introduce significant dose errors in individual field doses; however, these errors tend to cancel out between fields and average out over dose fractionation. 3D dose distributions, DVHs, TCPs, and EUDs for stationary and moving cases showed good agreement after two or more fractions, suggesting that tumors affected by respiration motion may be treated using IMRT without significant dosimetric and biological consequences. © 2006 American Association of Physicists in Medicine.