Despite recent attention given to the concept of modularity and its potential contribution to the evolvability of organisms, there has been little mention of how such a contribution may affect rates of diversification or how this would be assessed. A first key prediction is that lineages with relatively greater degrees of modularity in given traits should exhibit higher rates of diversification. Four general conditions for testing this prediction of the modular evolvability hypothesis are outlined here. The potential role of modularity as a deterministic factor in adaptive radiations is best examined by looking at historic patterns of diversification rather than just levels of extant diversity, the focus of most analyses of key innovations. Recent developmental evidence supports the notion that phenotypes of juvenile and adult stages of insects with "complete" metamorphosis (Holometabola) are distinct developmental and evolvable modules compared to the highly correlated life stages of insects with "incomplete" metamorphosis (Hemimetabola). Family-level rates of diversification for these two groups were calculated from the fossil record. The Holometabola was found to have a significantly and characteristically higher rate of diversification compared to the less modular Hemimetabola, consistent with the idea that intrinsic differences in modularity can influence the long-term evolvability of organisms. The modular evolvability hypothesis also makes a second key prediction: that characters in more modular clades will exhibit greater levels of variation due to their independence. This provides an independent, phenotypically based test of the hypothesis. We discuss here how this second prediction may be tested in the case of the Hemi- and Holometabola.