Links between the stratospheric thermal structure and the ozone distribution are explored in the Goddard Earth Observing System chemistry-climate model (CCM). Ozone and temperature fields are validated using estimates based on observations. An experimental strategy is used to explore sensitivities of temperature and ozone using the CCM alongside the underlying general circulation model (GCM) with ozone specified from either observations or from a chemistry-transport model (CTM), which uses the same chemical modules as the CCM. In the CTM, upper stratospheric ozone is biased low compared to observations; GCM experiments reveal that using CTM ozone reduces a warm temperature bias near the stratopause in the GCM, and this improvement is also seen in the CCM. Near 5 hPa, the global-mean ozone profile is biased low in the CTM but is close to observations in the CCM, which suggests that the temperature feedbacks are important in simulating the ozone distribution in the middle stratosphere. In the low stratosphere there is a high bias in simulated ozone, which forces a warm bias in the CCM. The high ozone also leads to an overestimate in total column ozone of several tens of Dobson units in the polar regions. In the late part of the twentieth century the seasonal activation of chlorine, especially over Antarctica, destroys ozone as expected, so that chlorine-induced ozone decreases are overestimated in the CCM compared to the real atmosphere. Ozone-change experiments reveal that the thermal structures of the GCM and CCM respond in a similar manner to ozone differences between 1980 and 2000, with a peak ozone-induced temperature change of about 1.5 K (over 20 years) near the stratopause, which is at the low end of the range computed by other models. Greenhouse-gas-induced cooling increases with altitude and, near the stratopause, contributes an additional 1.3 K to the cooling near 1 hPa between 1980 and 2000. In the Antarctic, the ozone hole is simulated with some success by the CCM. As with many other models, the polar vortex is too persistent in late winter, but counteracting this, the CCM undergoes too much midwinter variability, meaning the ozone hole is more variable than it is in the real atmosphere. Temperature decreases associated with the ozone hole in the CCM are similar to those computed with other models.