We investigated high-resolution simulations of regional climate models (RCMs) driven by ERA-40 reanalyses over areas of selected European countries (Austria, Czech Republic, Hungary, Slovakia and Romania) for the period 1961-1990. RCMs were run at a spatial resolution of 10 km in the framework of the CECILIA project, and their outputs were compared with the EOBS dataset of gridded observations and RCM simulations at coarser 25 km resolution from the ENSEMBLES project to identify a possible gain from the CECILIA experiments over ENSEMBLES. Cold biases of air temperature and wet biases of precipitation dominate in the CECILIA simulations. Spatial variability and distribution of the air temperature field are well captured. The precipitation field, relative to observations, often shows inadequately small spatial variability and lowered correlations but is nevertheless comparable to the ENSEMBLES model. Inter-annual variability (IAV) of air temperature is captured differently among seasons but mostly improved in CECILIA compared with ENSEMBLES. Precipitation IAV shows a similar or worse score. The detected weaknesses found within the validation of the CECILIA RCMs are attributed to the resolution dependence of the set of physical parameterizations in the models and the choice of integration domain. The gain obtained by using a high resolution over a small domain (as in CECILIA) relative to a lower resolution (25 km) over a larger domain (as in ENSEMBLES) is clear for air temperature but limited for precipitation.