The European Space Agency launched the Soil Moisture Ocean Salinity (SMOS) satellite in November 2009. We evaluated the reprocessed SMOS soil moisture retrievals for 2010 (V4.00). Assimilating satellite-derived soil moisture into initial conditions for weather forecast models will only improve short-term forecasts if the retrieval errors are small enough. Here we use in situ soil moisture observations to evaluate SMOS retrievals in a data-rich region of Oklahoma, USA. A major issue in comparing satellite data with in situ data is that there are few locations where more than one in situ station exists within a SMOS footprint, a region 40 km in diameter. A single station would not be representative of the soil moisture in the entire footprint due to the high spatial variability of precipitation, and hence soil moisture. The region around Stillwater, Oklahoma contains a large number of soil moisture stations within a small area and is the focal point of this study. The United States Climate Reference Network (USCRN), operated by the National Climatic Data Center, has two stations near Stillwater with three sensors at each station. The Oklahoma Mesonet also operates three stations in this area. The Cosmic-ray Soil Moisture Observing System also monitors soil moisture at a site near Stillwater. We find large variability in the in situ data, even between individual USCRN sensors that are only a few meters apart. Removing the mean and normalizing the data brings them into better agreement but they still contain substantial differences, making it difficult to fully assess the performance of SMOS. The differences in soil moisture between the individual in situ observations were as large as the differences between any of the individual observations and the SMOS retrievals. We found similar results for a second SMOS footprint near Stillwater that contained all of the same in situ stations. We also found no significant differences between ascending and descending passes of SMOS, and the reprocessed SMOS data have a dry bias when compared to the average of all of the in situ stations in the footprints.