Correct placement of radiation fields relative to patient anatomy is essential in radiotherapy in order to minimise serious side effects to reduce the probability of recurrence of the tumour. One way to determine patient setup accuracy is to analyse portal images obtained in the therapy beam distal to the patient. A field placement analysis (FPA) method has been developed for detailed evaluation of patient setup by comparing positions of corresponding radiation field edges in digitised simulator and portal images. A simulator image is matched to a portal image using similar anatomical landmarks in both images and mapping these landmarks against each other applying a least squares minimisation approach. Discrepancies between the simulator field edge (reference) and a portal field edge are determined by comparing the distances between the central axis of the beam and corresponding edge segments and the angles of these segments with a reference line. Uncertainties in these distances and angles are to a large extent determined by the magnification, rotation and translation procedure. Uncertainties due to the FPA method itself are of about 1.0 mm and 0.5 degrees in portal images of head and neck fields. These FPA uncertainties are in general smaller than the variations due to patient setups. Matching of simulator and portal images of lateral pelvic fields revealed larger uncertainties: 1.7 mm and 1.1 degrees. Setup variations in this kind of pelvic radiation field are usually also larger, and therefore meaningful results can be obtained with the new FPA method.