In this work, we present a study of the coseismic and post-seismic crustal deformation associated to the M_w 6.3, 2009 April 6 L'Aquila earthquake from the analysis of GPS displacement time-series. We use a principal component decomposition-based inversion method to study the space- and time-dependent evolution of slip on faults without any a priori assumption on the model used to characterize the temporal evolution of crustal deformation. The method adopted allows us to account for the initial post-seismic deformation in estimating the coseismic displacements, in a consistent manner for the whole GPS network. We use elastic dislocation theory and a least-squares procedure to invert for the slip distribution on the mainshock fault (Paganica fault) and a second fault (Campotosto fault), where a M_w 5.2 aftershock occurred on April 9. The geometries for these faults are obtained from a singular value decomposition of precisely relocated aftershocks. We find that the use of complex fault geometries is not justified by the GPS observations available. An inversion that accounts for post-seismic slip to occur on both the Paganica and Campotosto faults provides a better fit to the GPS time-series observations, than using only the Paganica fault segment, at a 95 per cent confidence level. Within our resolution, afterslip regions do not migrate over time and are localized on fault patches that are approximately complementary to those of coseismic slip. We find that the position of some relevant afterslip patches is different if the inversion is performed assuming a fixed rake or not. We estimate the parameter a – b of rate- and state-dependent friction on those fault regions accommodating afterslip that are robustly characterized in our inversions. We find values of the order of 10⁻³, which is near the transition from potentially unstable to nominally stable friction. These results are in agreement with laboratory measurements performed on typical rocks of the L'Aquila region.