The maximum expected ground motion in Greece is estimated for shallow earthquakes using a deterministic seismic hazard analysis (DSHA). In order to accomplish this analysis the input data include an homogeneous catalogue of earthquakes for the period 426 BC–2003, a seismogenic source model with representative focal mechanisms and a set of velocity models. Because of the discrete character of the earthquake catalogue and of errors in location of single seismic events, a smoothing algorithm is applied to the catalogue of the main shocks to get a spatially smoothed distribution of magnitude. Based on the selected input parameters synthetic seismograms for an upper frequency content of 1 Hz are computed on a grid of 0.2°×0.2°. The resultant horizontal components for displacement, velocity, acceleration and DGA (Design Ground Acceleration) are mapped. The maps which depict these results cannot be compared with previously published maps based on probabilistic methodologies as the latter were compiled for a mean return period of 476 years. Therefore, in order to validate our deterministic analysis, the final results are compared with PGA estimated from the maximum observed macroseismic intensity in Greece during the period 426 BC–2003. Since the results are obtained for point sources, with the frequency content scaled with moment magnitude, some sensitivity tests are performed to assess the influence of the finite extent of fault related to large events. Sensitivity tests are also performed to investigate the changes in the peak ground motion quantities when varying the crustal velocity models in some seismogenic areas. The ratios and the relative differences between the results obtained using different models are mapped and their mean value computed. The results highlight the importance in the deterministic approach of using good and reliable velocity models. © 2007 Elsevier B.V. All rights reserved.