On the basis of existing literature and newly collected empirical evidence, this article discusses the theoretical, empirical, research methodological and policy implications of a recently introduced knowledge taxonomy, the so-called Differentiated Knowledge Bases model (Asheim et al., 2011), which distinguishes between three epistemologically different approaches that are summarised in the notions of analytical (theoretically understanding), synthetic (instrumentally solving problems) and symbolic (culturally creating meanings). The article suggests that these differentiated knowledge bases, though ideal-typical constructs, seem applicable to micro-level, intra- or inter-organizational modes and communities of learning involved in firm innovation but that firms and meso- and macro-level social systems (sectors, clusters, regions, etc.) rarely rely on one single knowledge base but coordinate its actions in more learning modes and communities. The potentials for innovation research of this particular knowledge taxonomy are mainly connected with its integrative and wide perspective on the identification of the types of knowledge, modes of learning and institutional contexts that are relevant for firm innovation and regional economic development and that exceed the sectoral divides and production bias often characterizing innovation research. For innovation policy, this integrative perspective may provide new opportunities for encouraging the development, diffusion and use of economically valuable knowledge of different kinds and from varying societal spheres in ways that truly break with one-size-fits-all policies.