PURPOSE: To quantify the effect of set-up errors and respiratory motion on dose distributions for non-small cell lung cancer (NSCLC) treatment. METHODS AND MATERIALS: Irradiations of 5 NSCLC patients were planned with 3 techniques, two (conformal radiation therapy (CRT) and intensity modulated radiation therapy (IMRT1)) with a homogeneous dose in the planning target volume (PTV) and a third (IMRT2) with dose heterogeneity. Set-up errors were simulated for gross target volume (GTV) and organs at risk (OARs). For the GTV, the respiration was also simulated with a periodical motion around a varying average. Two configurations were studied for the breathing motion, to describe the situations of free-breathing (FB) and respiration-correlated (RC) CT scans, each with 2 amplitudes (5 and 10 mm), thus resulting in 4 scenarios (FB_5, FB_10, RC_5 and RC_10). Five thousand treatment courses were simulated, producing probability distributions for the dosimetric parameters. RESULTS: For CRT and IMRT1, RC_5, RC_10 and FB_5 were associated with a small degradation of the GTV coverage. IMRT2 with FB_10 showed the largest deterioration of the GTV dosimetric indices, reaching 7% for Dmin at the 95% probability level. Removing the systematic error due to the periodic breathing motion was advantageous for a 10 mm respiration amplitude. The estimated probability of radiation pneumonitis and acute complication for the esophagus showed limited sensitivity to geometrical uncertainties. Dmax in the spinal cord and the parameters predicting the risk of late esophageal toxicity were associated to a probability up to 50% of violating the dose tolerances. CONCLUSIONS: Simulating the effect of geometrical uncertainties on the individual patient plan should become part of the standard pre-treatment verification procedure.