We evaluate the ability of two coupled atmospheric-oceanic GCMs -the Hadley Center's third generation coupled climate model (HadCM3) and the Canadian Center for Climate Modeling and Analysis second-generation coupled model (CGCM2) -to simulate the North Atlantic Oscillation (NAO), the Pacific North American teleconnection pattern (PNA), and map patterns in the Midwest region of the United States, relative to NCEP/NCAR reanalysis (NNR) data. The observed (NNR-derived) and GCM-derived probability distributions and temporal behavior of the daily teleconnection indices exhibit agreement over the 1990-2001 reference period, and both GCMs successfully reproduce the range of 500-hPa map patterns over the study region. During the reference period, observed and modeled map patterns are similar in terms of frequency, coherence, persistence, and progression, although the most common map pattern occurs too often in HadCM3 relative to NNR and CGCM2-derived map patterns generally exhibit closer agreement with those derived from NNR data. Despite the relatively high degree of correspondence between the observed and simulated teleconnection indices and map patterns in the study area, differences between the GCM and NNR-derived map-pattern frequencies in the reference period are greater than either (1) recent historical changes in map-pattern frequencies or (2) changes in the mappattern frequencies as derived from twenty-first century GCM simulations, indicating that changes in these phenomena over recent and approaching decades are of insufficient magnitude relative to model uncertainty to be definitively identified.