Chemical transport models rely on meteorological analyses to provide temperatures which control model chemistry. An accurate representation of stratospheric temperatures becomes particularly important when temperatures at which polar stratospheric clouds (PSCs) are thought to form and chlorine activation begin are approached. This study examines the errors in United Kingdom Meteorological Office (UKMO) lower stratospheric temperatures during the winter 1994/95 by comparing them with temperatures obtained independently by ozonesondes during the same winter. The study shows that there is a consistently positive bias in the analysed temperatures which leads to a mean overestimation of lower stratospheric temperatures of up to 1.7 K at the NAT point. There is also a significant random scatter which increases with altitude. If corrections in analysis temperature are made for both bias and scatter, considerably larger areas of the Northern Hemisphere are exposed to temperatures below the NAT point and frost point and hence chemical activation. These discrepancies in the UKMO temperature fields have important implications for modeling studies carried out using these analyses.¿ 1997 American Geophysical Union