Several radiosonde (sonde) temperature datasets have been compiled and compared with two microwave-based temperature datasets to mutually test for inhomogeneities. In the case of the VIZ/Sippican sondes, the two satellite sets find several common inhomogeneities that correspond with known changes in sonde instrumentation or software. However, in some cases, the sonde dataset and one satellite dataset appear to detect an inhomogeneity in the other satellite dataset. For the VIZ sondes, in general, tropospheric trends are only slightly affected by the inhomogeneities while the stratospheric trends appear to be spuriously too negative. The Australian sondes show highly significant shifts to warmer tropospheric temperatures relative to both satellite datasets, indicating spurious warming. In a tropical dataset of 58 sondes it appears that the satellites detect inhomogeneities in some instances. However, the sondes also appear to detect problems in the satellite time series (about 0.1 K) at points where satellite adjustments are especially large and uncertain. This demonstrates that reference sondes , with traceable standards and consistent calibration, have the potential of calibrating bulk, atmospheric temperatures derived from microwave radiances.