Atmospheric temperature retrievals obtained by a simultaneous physical scheme and a statistical scheme are compared using data gathered over the Weddell Sea, Antarctica during January 1986. Ground truth data were obtained from radiosonde ascents. The statistical approach gave errors smaller than those of the physical method. Several modifications to the physical scheme were tried but, although some individual retrievals did show an improvement, the errors were still greater than for the statistical scheme. The results are for retrievals over the sea or low lying coast; it is expected that the physical scheme will give better results than the statistical scheme over the high Antarctic plateau as it allows for the effect of ground elevation in the temperature profile determination.