Three northern hemisphere winter/springs with contrasting thermal and dynamic characteristics were simulated with a mechanistic model of the middle atmosphere incorporating an interactive chemistry scheme. The three seasonal simulations used different planetary-wave forcings at the lower boundary, and consequently their respective temperature evolutions represented (1) a relatively mild Arctic winter; (2) a winter comparable to the coldest on record; and (3) a winter more persistently cold than any yet observed. The radiative response to the varying degrees of chemical ozone depletion in these winters was investigated by comparing results from runs that included and excluded the heterogeneous reactions on polar stratospheric clouds (PSCs) which initiate the enhanced destruction of polar ozone. Cooling of the lower stratosphere induced by the enhanced ozone loss ranges from virtually zero in the warmest winter to ~5 K in the coldest. The extent of any self-perpetuation of the ozone destruction processes via the induced cooling was examined by using the meteorological fields calculated both with and without PSC-initiated ozone depletion to drive an off-line chemical transport model. In winters 1 and 2 the longevity of PSCs and enhanced reactive halogen concentrations is increased only marginally by the cooling arising from the ozone loss, and consequently there is little impact on the net ozone destruction. In winter 3, however, the radiative feedback from the ozone loss leads to a marked prolongation of the lifetime of the enhanced reactive halogens and causes a chemically induced diminution of up to ~20 Dobson units (DU) in the April ozone column compared with the no-feedback case. Moreover, in winter 3 the cooling caused by PSC-initiated ozone depletion delays the springtime weakening of the polar vortex and inhibits the replenishment of Arctic ozone, resulting in a ~25 DU depression of the ozone column in early May that is due mainly to dynamics. These results suggest that chemical-radiative interaction via PSC processes has had only a small impact on the ozone distribution in recent Arctic winters, but that it has the potential to become more important should an unprecedentedly cold winter occur while atmospheric halogen concentrations remain artificially high. ¿ 2000 American Geophysical Union