This paper puts the Reinhart-Rogoff data-set to a formal econometric testing to see whether public debt has a negative non-linear effect on growth if public debt exceeds 90% of GDP. Using non-linear threshold models, we show that the negative non-linear relationship between debt and growth is very sensitive to modelling choices. We also show that when non-linearity is detected, the negative non-linear effect kicks in at much lower levels of public debt (between 20% and 60% of GDP). These results, based on bi-variate regressions on secular time series, are confirmed on a shorter data-set (1960-2010) using a multivariate growth framework.