The consistency of land-based and ocean-based Last Glacial Maximum (LGM) proxy evidence on and near South America is evaluated by utilizing a regional climate model (RCM). The relatively high resolution of the RCM brings model output closer to the spatial scales of the geological reconstructions. This is important because proxy sites only a few tens of kilometers apart can yield markedly different interpretations of LGM climate. Along with differences in orbital parameters, CO2 concentrations, and vegetation, four versions of LGM sea surface temperatures (SSTs) (Climate: Long-Range Investigation, Mapping, and Prediction (CLIMAP), Sch¿fer-Neth and Paul's (2003) core version, Paul and Sch¿fer-Neth's (2003) line version, and coupled GCM output from Shin et al. (2003)), are used to generate four realizations of the LGM climate in the model. Mean annual surface temperature, mean temperature of the coldest month, and aridity differences from a present-day simulation are compared with the geological reconstructions over South America to evaluate the consistency of the land and ocean-based reconstructions. Model results suggest that the tropical Atlantic during the LGM was not significantly cooler than CLIMAP's original estimates. Simulations utilizing tropical Atlantic LGM SSTs cooler than CLIMAP estimates yield unrealistically warm LGM surface conditions over the equatorial Amazon basin. Reductions in evaporation and the atmosphere's moisture content over too-cold tropical oceans lead to a reduction in cloud cover over the Amazon and increases in solar heating of the surface. All of the LGM SST simulations produce drier conditions over tropical South America; however, the extent of the aridity is associated with the degree of cooling in the surrounding oceans, especially the tropical Atlantic. The line version LGM SSTs of Paul and Sch¿fer-Neth (2003) are recommended.