We present a new gridded climate reconstruction for Europe for the last 12,000 years based on pollen data. The reconstruction is an update of Davis et al. (2003) using the same methodology, but with a greatly expanded fossil and surface-sample dataset and more rigorous quality-control. The modern pollen dataset has been increased by more than 80%, and the fossil pollen dataset by more than 50%, representing almost 60,000 individual pollen samples. The climate parameters reconstructed include summer/winter and annual temperatures and precipitation, as well as a measure of moisture balance, and growing degree-days above 5 degrees C. Confidence limits were established for the reconstruction based on transfer function and interpolation uncertainties. The reconstruction takes account of post-glacial isostatic readjustment which resulted in a potential warming bias of up to +1-2 degrees C for parts of Fennoscandia in the early Holocene, as well as changes in palaeogeography resulting from decaying ice sheets and rising post-glacial sea-levels. This new dataset has been evaluated against previously published independent quantitative climate reconstructions from a variety of archives on a site-by-site basis across Europe. The results of this comparison are generally very good; only chironomid-based reconstructions showed substantial differences with our values. Our reconstruction is available for download as gridded maps throughout the Holocene on a 1000-year time-step. The gridded format makes our reconstructions suitable for comparison with climate model output and for other applications such as vegetation and land-use modelling. Our new climate reconstruction suggests that warming in Europe during the mid-Holocene was greater in winter than in summer, an apparent paradox that is not consistent with current climate model simulations and traditional interpretations of Milankovitch theory. (C) 2015 Elsevier Ltd. All rights reserved.