Innovation heterogeneity refers to two empirical facts: economic sectors vary according to sources and rates of innovation, and innovations vary in terms of the magnitude of their economic impact. The central focus of this paper is the problem of scale effects in the Schumpeterian growth models. Although these models make endogenous the production of innovations, they assume not only an oversimplified pattern of sectoral innovation but also that major innovations are virtually indistinguishable from minor innovations. The main claim of the a er is that without a theoretical framework revolving around both the existence of realistic sectoral patterns of innovation and the explicit incorporation of major and minor innovations it is extremely difficult to resolve the scale effects problem in a satisfactory fashion. We disaggregate the 'ideas production function' in a way that may be useful to guiding future research aimed at ameliorating the intensity of the scale effects and introduce the concepts of 'awkward fact' and 'innovation regime'. A subsidiary claim is that our approach throws light on the ongoing controversy between neoclassical and evolutionary theorizing.