Bai et al. suggest that in China's Inner Mongolia steppe, community-level stability arises from compensatory effects among the principal components at both the species and plant functional group (PFG) levels. By analysing a consistent 19-year data set (1980-98), we show here that their analysis of a 24-year field data set (1980-2003) is called into question by inconsistencies in sampling location and numbers after 1998; the authors' findings are further undermined because they do not distinguish temporal variation from spatial heterogeneity in analysing compensatory effects among species or PFGs. We believe that rigorous reanalysis is needed for a better understanding of grassland stability.