The purpose of this study is to establish the in vivo verification of proton beam path by using proton-activated positron emission distributions. A total of 50 PET/CT imaging studies were performed on ten prostate cancer patients immediately after daily proton therapy treatment through a single lateral portal. The PET/CT and planning CT were registered by matching the pelvic bones, and the beam path of delivered protons was defined in vivo by the positron emission distribution seen only within the pelvic bones, referred to as the PET-defined beam path. Because of the patient position correction at each fraction, the marker-defined beam path, determined by the centroid of implanted markers seen in the posttreatment (post-Tx) CT, is used for the planned beam path. The angular variation and discordance between the PET- and marker-defined paths were derived to investigate the intrafraction prostate motion. For studies with large discordance, the relative location between the centroid and pelvic bones seen in the post-Tx CT was examined. The PET/CT studies are categorized for distinguishing the prostate motion that occurred before or after beam delivery. The post-PET CT was acquired after PET imaging to investigate prostate motion due to physiological changes during the extended PET acquisition. The less than 2 degrees of angular variation indicates that the patient roll was minimal within the immobilization device. Thirty of the 50 studies with small discordance, referred as good cases, show a consistent alignment between the field edges and the positron emission distributions from the entrance to the distal edge. For those good cases, average displacements are 0.6 and 1.3 mm along the anterior-posterior (D(AP)) and superior-inferior (D(SI)) directions, respectively, with 1.6 mm standard deviations in both directions. For the remaining 20 studies demonstrating a large discordance (more than 6 mm in either D(AP) or D(SI)), 13 studies, referred as motion-after-Tx cases, also show large misalignment between the field edge and the positron emission distribution in lipomatous tissues around the prostate. These motion-after-Tx cases correspond to patients with large changes in volume of rectal gas between the post-Tx and the post-PET CTs. The standard deviations for D(AP) and D(SI) are 5.0 and 3.0 mm, respectively, for these motion-after-Tx cases. The final seven studies, referred to as position-error cases, which had a large discordance but no misalignment, were found to have deviations of 4.6 and 3.6 mm in D(AP) and D(SI), respectively. The position-error cases correspond to a large discrepancy on the relative location between the centroid and pelvic bones seen in post-Tx CT and recorded x-ray radiographs. Systematic analyses of proton-activated positron emission distributions provide patient-specific information on prostate motion (sigmaM) and patient position variability (sigmap) during daily proton beam delivery. The less than 2 mm of displacement variations in the good cases indicates that population-based values of sigmap and sigmaM, used in margin algorithms for treatment planning at the authors' institution are valid for the majority of cases. However, a small fraction of PET/CT studies (approximately 14%) with -4 mm displacement variations may require different margins. Such data are useful in establishing patient-specific planning target volume margins.