The predictability of winter-time North Atlantic sea level pressure (SLP) variability has been investigated by means of an ensemble of integrations with an atmospheric general circulation model (AGCM) forced by observed sea surface temperatures (SSTs) for the period 1951--1994. The results imply that the SLP variations on timescales of several years to decades may be predictable, provided the SST anomalies themselves used to drive the AGCM can be predicted. The model, however, suffers from systematic errors, and the simulated centers of action are shifted relative to those observed. ¿ 2000 American Geophysical Union