Atmospheric carbon dioxide (CO2) estimates for the Middle Cretaceous (MK) have a range of >4000 ppm, which presents considerable uncertainty in understanding the possible causes of warmth for this interval. This paper examines the problem of MK greenhouse forcing from an inverse perspective: we estimate upper ocean water temperatures from oxygen isotope measurements of well-preserved late Albian--Turonian planktonic foraminifera and compare these against temperatures predicted by general circulation model (GCM) experiments with CO2 concentrations of 500--7500 ppm. At least 4500 ppm CO2 is required to match maximum temperatures inferred from well-preserved planktonic foraminifera. Approximately 900 ppm CO2 produces a good match between the model and the minimum temperature estimates for the MK. An ocean model forced by these two extremes in surface conditions brackets nearly all available bottom water temperature estimates for this interval. The climate model results support nearly the entire range of MK CO2 estimates from proxy data. The ocean model suggests possible MK oceanographic changes from deep water formation in the high latitude region of one hemisphere to the other hemisphere in response to changes in atmospheric temperatures and hydrologic cycle strength. We suggest that, rather than contradicting one another, the various proxy CO2 techniques (especially those with high temporal resolution) may capture true variability in CO2 concentrations and that MK CO2 could have varied by several thousand ppm through this interval.