The persistence of the Arctic and Antarctic lower stratospheric vortices is examined over the period from 1958 to 1999. Three different vortex-following diagnostics (two using potential vorticity and one based solely on the zonal winds) are compared and are shown to give very similar results for the breakup date. The variability in the timing of the breakup of both vortices is qualitatively the same: There are large interannual variations together with smaller decadal-scale variations and there is a significant increase in the persistence since the mid-1980s (all variations are larger for the Arctic vortex). Also, in both hemispheres, there is a high correlation between the persistence and the strength and coldness of the spring vortex, with all quantities having the same interannual and decadal variability. However, there is no such correlation between the persistence and the characteristics of the midwinter vortex. In the Northern Hemisphere, there is also a high correlation between the vortex persistence and the upper tropospheric/lower stratospheric eddy heat flux averaged over the 2 months prior to the breakup. This indicates that the variability in the wave activity entering the stratosphere over late winter to early spring plays a key role in the variability of the Arctic vortex persistence (and spring polar temperatures) on both interannual and decadal timescales. However, the extreme values of Arctic vortex coldness and persistence during the 1990's are not echoed as a similar extreme in the eddy heat flux. This suggests that the recent increase in vortex persistence is not solely due to changes in the wave activity entering the stratosphere. ¿ 1999 American Geophysical Union