"This paper marks a departure in seeking to develop the conceptual and practical apparatus of a regional innovation system (RIS) for science & technologydisadvantaged regions. It is empirically based and builds on insights about the limitations of STI (The Science-Technology-Innovation Approach, which is Linear, Specialist, Exclusive, Explicit/Codified, Global) and the strengths of DUI (The Doing-Using-Interacting Approach, which is Interactive, Diversified, Inclusive, Implicit, Regional/Local). DUI is highly compatible with Schumpeterian understanding that the core process of innovation is 'knowledge recombination'. From an evolutionary economic geography perspective, which is taken in the paper, this raises interesting issues for the economics of knowledge. First it underlines the need to pay serious attention to questions of the 'proximity' imperative, suggesting not that knowledge is easily appropriable for ('open') innovation but that it may be excessively difficult to identify because it lies hidden in possibly neighbouring - but different - industries and firms. Thus, second, it makes the notion of 'knowledge spillovers' problematic because the spillovers may not be forthcoming at all or may come in unrecognisable forms. Hence, third, this means that firms likely need more than usual RIS intermediation (including knowledge demonstration and transfer services) to avoid market failures of innovation. Assistance with identification of ‘modular’ policy elements is only one of the services required for DUI product, process and policy innovation. The complexity theory notion of 'transversality' has been advanced to capture the 'emergence' of novelty out of contexts of difference, unifying a solution to the three conceptual problem-issues raised in the paper."