Evaluations of operational TIROS-N and NOAA-6 temperature soundings over North America are presented for an early January 1980 period one month after completion of the First GARP Global Experiment. In addition to collocated comparisons, synoptic analyses derived only from satellite data and model forecasts initialized from these analyses are compared with those obtained from conventional data. The collocated results, similar to those presented by Phillips et al. (1979) and Schlatter (1981) from TIROS-N soundings, show maximum sounding errors new the surface and tropopause. The analysis comparisons further illustrate that thermal gradients inferred from satellite soundings are too weak, with NOAA-6 gradients slightly weaker than TIROS-N gradients. Difference fields between satellite and conventional thickness analyses propagate eastward with the synoptic patterns, strongly suggesting a correlation of satellite sounding errors to synoptic patterns. These anomalies are also retained in model forecasts started from satellite analyses. These results stress the importance of properly defining the error characteristics of satellite soundings before incorporating them into analysis-forecast systems.