The economics profession seems to increasingly endorse the existence of a strongly negative nonlinear effect of public debt on economic growth. Reinhart and Rogoff (2010) were the first to point out that a public debt-to-GDP ratio higher than 90% of GDP is associated with considerably lower economic performance in advanced and emerging economies alike. A string of recent empirical papers broadly validates this threshold value. This paper seeks to contribute to this literature by putting a variant of the Reinhart-Rogoff dataset to a formal econometric testing. Using nonlinear threshold models, there is some evidence in favour of a negative nonlinear relationship between debt and growth. But these results are very sensitive to the time dimension and country coverage considered, data frequency (annual data vs. multi-year averages) and assumptions on the minimum number of observations required in each nonlinear regime. We show that when non-linearity is detected, the negative nonlinear effect kicks in at much lower levels of public debt (between 20% and 60% of GDP). These results, based on bivariate regressions on secular time series, are largely confirmed on a shorter dataset (1960-2010) when using a multivariate growth framework that accounts for traditional drivers of long-term economic growth and model uncertainty. Nonlinear effects might be more complex and difficult to model than previously thought. Instability might be a result of nonlinear effects changing over time, across countries and economic conditions. Further research is certainly needed to fully understand the link between public debt and growth.