Two of the largest historical earthquakes documented for the United States occurred within the past four decades in a zone of anomalous seismicity in the western interior of the country known as the Intermountain Seismic Belt (ISB). Paleoseismology has revealed the presence of major active faults throughout this region, some of them extending through densely populated areas. This study presents mean-value estimates of the maximum earthquake magnitudes that can be expected from 65 of the most prominent late Quaternary fault segments of the ISB using surface rupture parameters. Linear, least-squares regressions of magnitude on surface rupture length, L, maximum surface displacement, D, and DL, respectively, were computed for an updated worldwide set of historical, normal-slip earthquakes. These were compared with similar regressions from other publications, and it was found that a DL-based formula restricted to normal-faulting mechanisms has a relatively high correlation, low standard deviation, and tends to yield magnitudes between those given by D and L alone. This last property is useful for applications to the ISB prehistoric data, where the record of historical, ground-rupturing earthquakes is too sparse to estimate the proportions in which D and L are likely to contribute to seismic energy release of a fault. Application of regressions to published paleoseismic displacements and estimated earthquake segment lengths for the ISB faults shows that D consistently scales a larger magnitude than the corresponding single-segment L by an average of 0.2 to 0.3 magnitude unit, after accounting for bias from the regressions. Considerations of uncertainties in paleoseismic displacements suggest that they cannot fully explain the discrepancy. Since it is known that large earthquakes often encompass multiple segments of a fault, these observations indicate that longer rupture lengths should be used to estimate maximum magnitudes for ISB faults. A dual-segment rupture scenario produces better agreement between D- and L-based magnitudes and is supported by historical and paleoseismic earthquake data from the region. Magnitudes scaled by the DL-based regression for dual-segment rupture range from Ms 6.8 to 7.5 (m = +/-0.20), and the largest are associated with the central segments of the Wasatch fault and the Swan/Grand Valley, Teton, Madison, and East Bear Lake faults.