Coal rank reflects the temperature of coalification, with higher rank coal forming at higher temperatures. If the temperature of coalification depended only on the Earth's geothermal gradient, then the maximum rank of coal in a sedimentary basin should be directly proportional to the thickness of strata above the coal. This association does not occur in the Illinois Basin, a continental-interior basin in the midwestern United States, where the overall coal rank observed is higher than can be explained by past burial depth alone. Recognition of this anomaly has led many authors to suggest that hot groundwater flowing from south to north, during a Paleozoic basin-scale groundwater-migration event, increased coal rank. We analyzed vitrinite reflectance (R-o), a measure of rank, as a function of depth in wells across the basin to determine how paleogeotherms, representing variation in temperature with depth, change with location. Our results show that the basin can be divided into three zones: (1) in the southern zone, the paleogeotherm varies irregularly with depth in strata above the sub-Absaroka unconformity (the contact separating Pennsylvanian and Mississippian strata); (2) in the central zone, the paleogeotherm displays a distinct inflection at the unconformity, for the rate of increase in Ro with depth is greater above the unconformity than below the unconformity; and (3) the observed inflection in the paleogeotherm dies out northward, until, in the northern zone, the paleogeotherm has the same slope both above and below the unconformity. We propose that the inflection in the paleogeotherm of the central zone indicates that the hot groundwater responsible for causing an increase in coal rank flowed through a high-permeability zone just below the sub-Absaroka unconfoimity. This flow, which advected heat into Pennsylvanian strata above, cooled as it moved northward, so it did not influence the paleogeotherm in the northern zone. Preliminary studies of vein paragenesis, stable-isotope composition, and fluid-inclusions, as well as computer modeling of flow-related heat advection, support this proposal.