This paper deals with the debt-growth relationship using several time-series tools. The idea is to find out whether the inverse relationship between these variable can be detected without imposing any functional forms for the estimating relationship and whether the relationship does indeed reflect some nonlinear features. Thus recursive correlations with different orderings of the time-series are computed using the Reinhart and Rogoff panel data. After that, recursive correlations are re-estimated with data that are cyclically adjusted to reflect the structural features of these two variables. The nature of the relationship is also scrutinized by using various variable-parameter estimation techniques (Kalman Filter, Logistic functional form and recursive estimation). Finally, some analyses of causality are carried out using these (filtered) data. The analysis clearly shows that the inverse relationship is very robust indeed and it rather supports the "toxic debt" hypothesis than the cyclical debt accumulation hypothesis.