This article discusses the implications for competition, innovation and learning of different forms of inter-firm linkage, ways to govern them, different 'generic systems' of innovation, and government policy. It employs a transformed theory of transactions that can deal with innovation and learning, and brings in trust next to opportunism [Nooteboom, B., 1996a. Trust, opportunism and governance: a process and control model. Organization Studies 17 (6) 985-1010; Nooteboom, B., 1996b. Towards a Learning Based Model of Transactions. In: Groenewegen, J. (Ed.), TCE and Beyond. Kluwer, Deventer, pp. 327-349; Nooteboom, B., 1999a. Inter-firm alliances: Analysis and design. Routledge, London.]. While trust has its limits and should not be blind, it can lower transaction costs. For learning and innovation, it takes the resource/competence perspective, supported by a theory of knowledge developed in earlier publications. According to this theory people perceive, interpret and evaluate the world according to cognitive categories that have developed in interaction with the physical and social environment. As a result people will perceive, understand and evaluate differently to the extent that they have developed in different environments without interaction [Nooteboom, B., 1992,. Towards a dynamic theory of transactions. Journal of Evolutionary Economics 2, 281-299; Nooteboom, 1999a.]. This theory yields the notion of 'external economy of cognitive scope': people and firms need outside sources of cognition and competence to complement their own. That is the fundamental reason why inter-firm Linkages are important, especially for innovation. In order to produce high added value and novelty, by utilizing the opportunities of complementary competencies, firms need to make relation-specific investments which creates risks of 'hold-up' and 'spill-over'. Building on earlier work, the article identifies different instruments for the control of those risks [Nooteboom, 1996a; Nooteboom, 1996b; Nooteboom, et al., 1997. Effects of trust and governance on relational risk. Academy of Management Journal 40 (2) 308-338; Nooteboom, 1999a.]. It identifies two 'generic' kinds of innovation systems, in terms of the mix of instruments for relational governance, and discusses their merits and flaws with respect to quality of products, diffusion, incremental and radical innovation. One is close to practices in continental Europe and Japan. Another is close to Anglo-American practice. There is a certain tendency for the first to gravitate to the second. The article warns about the dangers involved, and explores a possible 'third way'. (C) 1999 Elsevier Science B.V. All rights reserved.