The reliability of the recently published reconstructions of the surface air temperature variability in the Northern Hemisphere over the past 2000 yr is discussed. For this purpose, the power spectra of the two best known reconstructions (Mann et al.[10–12] and Moberg et al. [13]) are calculated and compared to the spectra of the 150-yr temperature series based on instrumental observations and simulated 1000-yr series. It is found that the Mann et al. reconstruction drastically underestimates low-frequency temperature variations, whereas the Moberg et al. reconstruction reproduces them much better, although with a certain underestimation rather than overestimation, as Mann et al. have recently argued.