Patterns of temperature variability on interdecadal and longer timescales are derived from empirical and simulated data for the region centered around the North Atlantic ocean. The empirical dataset consists of long early-instrumental records and high-quality proxy data. The simulated data is obtained from a long integration of the coupled intermediate-complexity model ECBilt. A combination of classical statistical tools and modern spectral methods is used in the analysis. Both in the empirical and simulated data, temperature variability is found to be season dependent. In the empirial data, two distinct statistically significant spatio-temporal modes of temperature variability are identified on timescales longer than 50 years. The simulated data do not exhibit any dominant timescale of variability in the considered range, which might be due to the absence of external periodic forcing in the model experiment. In summer, there is a promising similarity between empirical and simulated patterns of temperature variability.