Empirical evaluation of the magnitude-independence assumption