Temperature histories for selected stratigraphic horizons in the Michigan basin are computed from three-dimensional continually filled subsidence models. Mechanical evolution of the Michigan basin is modeled as flexure of the lithosphere caused by thermal contraction. Results are compatible with the subsidence record of the sediments and free-air gravity anomalies. Paleotemperature is determined from excess temperature due to the thermal anomaly plus burial temperature predicted from subsidence curves. For an equilibrium temperature gradient of 22°C/km (1.2°F/100 ft), surface temperature of 20°C (68°F), and an equilibrium surface heat flow of 1.1 HFU, excess temperature, paleotemperature, and surface heat flow do not exceed 15°C (27°F), 110 76;C (230°F), and 2.5 HFU, respectively. These estimates are consistent with upper limits set by paleomagnetic studies. The low value for excess temperature is caused by concentration of the thermal anomaly below 15 km (9 mi), in agreement with gravity results. The great depth of the thermal anomaly can explain the lack of evidence for an initial heating event prior to subsidence. Once the thermal history of the sediments is specified, the oil potential of the basin can be determined from laboratory-derived kinetic equations for degradation of kerogen to petroleum. For the Michigan basin, predicted temperature conditions are sufficient for source rock (> 25%) conversion of type I and type II kerogen only in Ordovician and older rocks in the southern peninsula of Michigan. By this model, petroleum found in rocks younger than Ordovician would have had to migrate upward from the older rocks. Geochemical studies of Dundee (Devonian) and Trenton (Ordovician) crude oils in Michigan are compatible with this interpretation. Niagaran (Silurian) crude oils appear to come from a different source than the Dundee and Trenton oils. If their source is Silurian rocks on the flanks of the basin, either the scheme used to calculate maturity is not applicable to these oils or the temperature of the source rocks was significantly higher in the past than today. More likely, the model underestimates the depth of burial of Silurian source rocks near the center of the basin, especially in local regions of downfaulting.