We investigate the uncertainty of estimated peak ground acceleration (PGA), 0.2 s spectral acceleration (Sa), and 1.0 s Sa to different input variables for the southern Illinois basin (SIB) using the model and computer codes of the 1996 U.S. Geological Survey (USGS) National Seismic Hazard maps of Frankel et al. (1997b). The uncertainties associated with incomplete knowledge are assessed using a Monte Carlo approach to sample a logic tree. The logic tree is made of input-variable branch points representing alternative values for each parameter. This study is the first whole-map uncertainty analysis for the smoothed-seismicity method on which much of the central and eastern U.S. National Seismic Hazard maps are based. The uniqueness of the approach used is the assessment of individual branch-point sensitivity on the uncertainty in the seismic hazard calculations, in addition to overall uncertainty. The mean 2%-probability-of-exceedence-in-50-year PGA, 0.2 s Sa, and 1.0 s Sa hazard maps from the logic tree are similar to the corresponding maps of Frankel et al. (1997b). Overall, the coefficient of variation (standard deviation divided by the mean for each grid point) in these maps varies between 0.2 and 0.7. The most sensitive branch point of the SIB logic tree is for the lower-magnitude limit of the catalog used in calculating seismic hazard. It is followed closely in importance by the branch points representing uncertainty in activity rate and the choice of ground-motion attenuation relation. Results suggest that the greatest reduction in uncertainty in seismic hazard calculations for the SIB could be achieved by continued research in the areas of (1) eastern U.S. earthquake monitoring (to extend the short instrumental catalog of earthquakes), (2) studies leading to paleoseismic information (to extend the historical catalog for large events), and (3) strong ground-motion studies.