We report the results of a prospective study to quantify inaccuracies in patient set-up during routine radiotherapy for tumours of the brain and bladder, which took place as part of our departmental development. Knowledge of these inaccuracies is required to put into practice the ICRU 50 recommendations regarding clinical target volume and planning target volume. We measured inaccuracies in two dimensions by comparing portal beam films with the simulator check film. Our method used manual measurements, proved to be a very laborious technique, and demonstrated the need for portal imaging. Ninety-five brain and 97 bladder portal films from 30 brain and 30 bladder patients were examined. Displacements greater than 0.5 cm were seen in 13% of brain treatments in the supero-inferior direction and 1% in the anteroposterior direction. With bladder treatments, displacements greater than 0.75 cm were seen in 12% in the supero-inferior direction and 5% in the lateral direction. These results are consistent with other previous studies. We identified a very small systematic error in the department, which was not [corrected] considered to be clinically significant. These results are discussed with reference to other similar studies and the ICRU 50 recommendations.