Recent literature suggests that long-run averages of growth and inflation are only weakly correlated and such correlation is not robust to exclusion of extreme inflation observations; inclusion of time series panel data has improved matters, but an aggregate parametric approach remains inconclusive. We propose a nonparametric definition of high inflation crises as periods when inflation is above 40 percent annually. Excluding countries with high inflation crises, we find no evidence of any consistent relationship between growth and inflation at any frequency. However, we find that growth falls sharply during discrete high inflation crises, then recovers surprisingly strongly after inflation falls. The fall in growth during crisis and recovery of growth after crisis tend to average out to close to zero (even slightly above zero), hence the lack of a robust cross-section correlation. Our findings could be consistent either with trend stationarity of output, in which inflation crises are purely cyclical phenomena, or with models in which crises have a favorable long-run purgative effect. Our findings do not support the view that reduction of high inflation carries heavy short-to-medium run output costs.