An ensemble of statistical methods is applied to reconstruct European temperature variability back to 1500. We apply principal component (PC) regression, regularized expectation maximization (RegEM) and composite-plus-scaling (CPS) to multi-proxy data. The reconstructions for summer and winter European temperature averages, and spatial fields related to warmest and coldest decades are analyzed and discussed. PC regression and RegEM perform more similarly compared to CPS, and more robust reconstructions are achieved for winter than for summer. We conclude that temperature reconstructions can not be improved significantly by replacing the reconstruction technique only. Discordances are very likely to be due to limited spatial and temporal availability of the proxy data. The comparison reveals that seasonal temperature variability is likely more variable than indicated earlier, still pointing out the exceptional warmth of the late 20th century. However, further evidence is needed, as the summer reconstruction results of the three techniques are not yet fully coherent.