Abstract More than 40 years of radiosonde data from two Antarctic stations are examined for changes in the date of the final stratospheric warming that occurs each year as the vortex breaks up in spring/summer. A new measure of this date is derived that does not rely on specification of a threshold, as has been common previously. The date of final warming takes between 10 and 40 days to progress from 30 to 100 hPa and occurs 20–30 days later in the 1990s than in the 1960s. Multiple linear regression analyses of these final warming dates, and also of the vertical profile of the southern annular mode (SAM), are presented. Only a weak signal is found for a linear trend, but a significant response is found throughout the atmosphere to ozone mass deficit (OMD), representing stratospheric ozone loss. In the SAM a significant response to the combined influence of solar variability and the quasi-biennial oscillation (QBO) is also found. The seasonal evolution of these signals in the NCEP Reanalysis zonal mean temperatures is examined and their influences on final warming dates is calculated. This confirms that ozone loss is primarily responsible for the delayed warming in the lower stratosphere in recent years, but suggests that the phase of solar activity and the QBO also have an effect. The apparent downward progression of the signal of OMD, and of the combined solar activity and QBO, extends well into the upper troposphere and appears to be caused simply by a delay in the top-down breakup of the vortex.