The purpose of this work was to simulate with the Monte Carlo (MC) code PENELOPE the dose distribution in lung tumours including breathing motion in stereotactic body radiation therapy (SBRT). Two phantoms were modelled to simulate a pentagonal cross section with chestwall (unit density), lung (density 0.3 g cm(-3)) and two spherical tumours (unit density) of diameters respectively of 2 cm and 5 cm. The phase-space files (PSF) of four different SBRT field sizes of 6 MV from a Varian accelerator were calculated and used as beam sources to obtain both dose profiles and dose-volume histograms (DVHs) in different volumes of interest. Dose distributions were simulated for five beams impinging on the phantom. The simulations were conducted both for the static case and including the influence of respiratory motion. To reproduce the effect of breathing motion different simulations were performed keeping the beam fixed and displacing the phantom geometry in chosen positions in the cranial and caudal and left-right directions. The final result was obtained by combining the different position with two motion patterns. The MC results were compared with those obtained with three commercial treatment planning systems (TPSs), two based on the pencil beam (PB) algorithm, the TMS-HELAX (Nucletron, Sweden) and Eclipse (Varian Medical System, Palo Alto, CA), and one based on the collapsed cone algorithm (CC), Pinnacle(3) (Philips). Some calculations were also carried out with the analytical anisotropic algorithm (AAA) in the Eclipse system. All calculations with the TPSs were performed without simulated breathing motion, according to clinical practice. In order to compare all the TPSs and MC an absolute dose calibration in Gy/MU was performed. The analysis shows that the dose (Gy/MU) in the central part of the gross tumour volume (GTV) is calculated for both tumour sizes with an accuracy of 2-3% with PB and CC algorithms, compared to MC. At the periphery of the GTV the TPSs overestimate the dose up to 10%, while in the lung tissue close to the GTV PB algorithms overestimate the dose and the CC underestimates it. When clinically relevant breathing motions are included in the MC simulations, the static calculations with the TPSs still give a relatively accurate estimate of the dose in the GTV. On the other hand, the dose at the periphery of the GTV is overestimated, compared to the static case.