Interannual variability in an ensemble of six 47-year climate simulations with specified time-varying sea surface temperature (SST) and sea ice extent is analyzed. Variability is compared with that in a 250-year control simulation with climatological SST and sea ice extent. The simulations were performed with the Canadian Climate Centre second generation General Circulation Model. An analysis of variance approach was combined with rotated Empirical Orthogonal Function analysis to assess the influence of the prescribed boundary conditions on the simulated atmospheric circulation structure and variability and to identify potentially predictable spatially coherent modes of variation. The quantities analyzed are seasonal mean 500 hPa geopotential (Z500) and mean sea level pressure (Pmsl). The prescribed boundary conditions increase the variability of the simulated atmosphere and modulate its large-scale circulation structure. The effects on Z500 are found to be strongest in DJF and MAM and weakest in SON. The prescribed boundary conditions have a significant effect over the extratropical land areas, especially over northern North America, in El Niño-Southern Oscillation (ENSO) years, but have little influence in non-ENSO years. The variability of several of the leading spatial modes is significantly affected by the prescribed boundary conditions. In contrast with some previous studies, we find that the simulated North Atlantic Oscillation is not affected by the prescribed boundary forcing.