I analyze the sensitivity of seismic-hazard estimates in the central and eastern United States (CEUS) to maximum magnitude (m_max) by exercising the U.S. Geological Survey (USGS) probabilistic hazard model with several m_max alternatives. Seismicity-based sources control the hazard in most of the CEUS, but data seldom provide an objective basis for estimating m_max. The USGS uses preferred m_max values of moment magnitude 7.0 and 7.5 for the CEUS craton and extended margin, respectively, derived from data in stable continental regions worldwide. Other approaches, for example analysis of local seismicity or judgment about a source's seismogenic potential, often lead to much smaller m_max. Alternative models span the m_max ranges from the 1980s Electric Power Research Institute/Seismicity Owners Group (EPRI/SOG) analysis. Results are presented as haz-ard ratios relative to the USGS national seismic hazard maps. One alternative model specifies m_max equal to moment magnitude 5.0 and 5.5 for the craton and margin, respectively, similar to EPRI/SOG for some sources. For 2% probability of exceedance in 50 years (about 0.0004 annual probability), the strong m_max truncation produces hazard ratios equal to 0.35-0.60 for 0.2-sec spectral acceleration, and 0.15-0.35 for 1.0-sec spectral acceleration. Hazard-controlling earthquakes interact with m_max in complex ways. There is a relatively weak dependence on probability level: hazardratios increase 0-15% for 0.002 annual exceedance probability and decrease 5-25% for 0.00001 annual exceedance probability. Although differences at some sites are tempered when faults are added, m_max clearly accounts for some of the discrepancies that are seen in comparisons between USGS-based and EPRI/SOG-based hazard results.