Stabilizing pollution levels in the long run is a pre-requisite for sustainable growth. We develop a neoclassical growth model with endogenous emission reduction predicting that, along optimal sustainable paths, pollution growth rates are (i) positively related to output growth (scale effect) and (ii) negatively related to emission levels (defensive effect). This dynamic law reduces to a convergence equation that is empirically tested for two major and regulated air pollutants - sulfur oxides and nitrogen oxides - with a panel of 25 European countries spanning the years 1980-2005. Traditional parametric models are rejected by the data. More flexible regression techniques confirm the existence of both the scale and the defensive effect, supporting the model predictions.