To quantify prostate motion during a radiation therapy treatment using cine-magnetic resonance imaging (cine-MRI) for time frames comparable to that expected in an image-guided radiation therapy treatment session (20-30 min).Six patients undergoing radiation therapy for prostate cancer were imaged on 3 days, over the course of therapy (Weeks 1, 3, and 5). Four hundred images were acquired during the 1-h MRI session in 3 sagittal planes through the prostate at 6-s intervals. Eleven anatomic points of interest (POIs) have been used to characterize prostate/bony pelvis/abdominal wall displacement. Motion traces and standard deviation for each of the 11 POIs have been determined. The probability of displacement over time has also been calculated.Patients were divided into 2 groups according to rectal filling status: full vs. empty rectum. The displacement of POIs (standard deviation) ranged from 0.98 to 1.72 mm for the full-rectum group and from 0.68 to 1.04 mm for the empty-rectum group. The low standard deviations in position (2 mm or less) would suggest that these excursions have a low frequency of occurrence. The most sensitive prostate POI to rectal wall motion was the mid-posterior with a standard deviation of 1.72 mm in the full-rectum group vs. 0.79 mm in the empty-rectum group (p = 0.0001). This POI has a 10% probability of moving more than 3 mm in a time frame of approximately 1 min if the rectum is full vs. approximately 20 min if the rectum is empty.Motion of the prostate and seminal vesicles during a time frame similar to a standard treatment session is reduced compared to that reported in interfraction studies. The most significant predictor for intrafraction prostate motion is the status of rectal filling. A prostate displacement of <3 mm (90%) can be expected for the 20 min after the moment of initial imaging for patients with an empty rectum. This is not the case for patients presenting with full rectum. The determination of appropriate intrafraction margins in radiation therapy to accommodate the time-dependent uncertainty in positional targeting is a topic of ongoing investigations for the on-line image guidance model.