This paper questions the line of reasoning followed by several authors, notably Easterly and Levine according to which ethno-linguistic fragmentation, because it leads to poor policies, is the main factor explaining the 'tragedy' of low African growth. A first set of criticism concerns the model itself and stresses that current empirical work is unable to convincingly identify the channels through which ethnic fragmentation affects growth: (i) polarization may be more relevant than fragmentation, (ii) the various tests of the effect of ethnicity on the quality of policy are far from being conclusive. A second set of remarks concerns the relevance of these studies to Africa: the African sub-sample is often quite limited, and the relationship is unstable (according to Chow tests). It actually appears that ethnicity has a more important effect on growth in Africa than elsewhere. This still needs to be explained and is not as such an explanation for lower African growth.