The influence of patient immobilization error on 3D planned conformal radiation therapy in tumors of the thorax and pelvis was studied. The mean positioning error in 43 patients with carcinomas of the thorax and pelvis undergoing 3D conformal radiotherapy (laser supported alignment, no immobilization device) was measured. A total of 194 portal films were superposed with the corresponding simulator radiographs according to anatomic landmarks and using a subtrascope. x-, y- and z-axis deviation was determined within a coordinate system. Using specialized software including Fourier transformation the mean positioning error was employed to recalculate the dose distributions of all cases under the influence of random (Gaussian) immobilization uncertainty. The mean two-dimensional positioning error using the data from all patients was 5.5 (+/- 3.7) mm. The distribution was Gaussian. Dose volume histograms (DVHs) of each patient with and without consideration of positioning uncertainty were compared on the base of tumor control probability estimations (TCP) using published DVH reduction and TCP algorithms. Inclusion of the positioning error resulted in a mean decrease in TCP (given as the difference between the TCP assuming no positioning error and the TCP modified by the positioning error) of 2% in a series of esophagus carcinomas and of 5% in the prostate carcinomas when looking at gross tumor volume (GTV), only. Planning target volume (PTV) exhibited a relative decrease in TCP of 5% and 11%, respectively.