Analysis of global surface air temperature records has indicated that recent years have been among the warmest since the late nineteenth century (Houghton et al., 1995); with 1995 being the warmest year on record (Hansen et al., 1996). But the rate of global annual mean surface warming of 0.13 deg C per decade during the period 1979-95 differs substantially from the global lower-tropospheric cooling trend of -0.05 \deg C per decade (Christy et al., 1995) inferred from the record (MSU-2R) of radiance measurements by the satellite Microwave Sounder Unit (MSU) (Spencer et al., 1990; Spencer and Christy, 1992). Accordingly, the satellite record has been widely cited by sceptics as evidence against global warming (Economist 1996; Michaels, 1996: Michaels, 1996; Singer, 1996; Spencer, 1996). However, a substantial fraction of the measured radiance originates not from the atmosphere but from the Earth's surface (Shah and Rind, 1995), and gives rise to high noise levels. This noise can lead to errors when merging temperature time series obtained from different satellites. Here we present comparisons among different MSU retrievals, sea surface temperatures (SSTs),and equivalent MSU temperatures derived from an atmospheric general circulation model forced with observed SSTs. The comparisons, focused on the tropics where atmospheric temperatures are closely tied to SSTs, strongly suggest that two spurious downward jumps occur in the MSU-2R record coinciding with changes in satellites, and that the real trend in MSU temperatures is likely to be positive, albeit small.