This study evaluates the dependence of simulated surface air temperatures on model resolution and orography for the mid-Holocene. Sensitivity experiments with the atmospheric general circulation model ECHAM5 are performed with low (∼3.75°, 19 vertical levels) and high (∼1.1°, 31 vertical levels) resolution. Results are compared to the respective preindustrial runs. It is found that the large-scale temperature anomalies for the mid-Holocene (compared to preindustrial) are significantly different in the low- and high-resolution versions. For boreal winter, differences are mainly related to circulation changes caused by the response to thermal forcing in conjunction with orographic resolution. For summer, shortwave cloud radiative forcing emerges as an important factor. The anomaly differences (low minus high resolution version) in the Northern Hemisphere are regionally as large as the anomalous mid-Holocene temperature signals. Furthermore, they depend on the applied surface boundary conditions. We conclude that the resolution matters for the Northern Hemisphere response in mid-Holocene simulations, which should be taken into account in model-model and data-model comparisons.