This paper seeks to gain insights on the relationship between growth and unemployment when considering heterogeneous agents in terms of skills. We allow for the possibility of train- ing for unskilled employed workers and for the possibility of human capital depreciation for skilled unemployed workers. These features are introduced in an endogenous job destruction framework µa la Mortensen and Pissarides (1998). We show that, when growth accelerates, a larger share of unskilled workers gets trained, increasing the incentives of ¯rms to update the job-speci¯c technology, rather than destroying it. The positive impact of growth on the employment rate is then magni¯ed and the predicting ability of the model to reproduce the sensibility of employment with respect to growth too. When calibrated, the model manages to reproduce the aggregate capitalization e®ect estimated on the basis of OECD data. Fur- thermore, whereas for skilled and unskilled workers getting trained growth yields a reduction in the unemployment rates, for unskilled workers not getting trained growth fosters a rise in the unemployment rates.