AbstractAssessing stratosphere–troposphere coupling in observational data or model output requires a multi‐level index with high time resolution. Ideally, such an index would (1) represent spatial patterns in the troposphere that are most strongly coupled with stratospheric variability and (2) be robust and computationally feasible in both observations and standard model output.Several of the indices used to diagnose extratropical stratosphere–troposphere coupling are based on the Northern and Southern Hemisphere annular modes. The annular mode indices are commonly defined as the leading empirical orthogonal functions (EOFs) of monthly‐mean, hemispheric geopotential height. In the lowermost troposphere, the structure of the annular modes is defined as the leading EOF of the near‐surface geopotential height field, and these patterns correspond well to the patterns of variability induced by stratospheric circulation changes. At pressure levels above the surface, the structure of the annular modes is typically found by either calculating the local EOF or regressing geopotential height data onto the leading principal component time series of near‐surface geopotential height.Here we make a critical comparison of the existing methodologies used to diagnose stratosphere–troposphere coupling, including EOF‐based indices as well as measures based on zonal‐mean wind at a fixed latitude and geopotential height over the polar cap. We argue in favour of an alternative methodology based on EOFs of daily zonally‐averaged geopotential. We find that (1) the daily evolution of stratosphere–troposphere coupling events is seen most clearly with this methodology, and (2) the methodology is robust and requires few subjective choices, making it readily applicable to climate model output available only in zonal‐mean form. Copyright © 2009 Royal Meteorological Society