Governance failures are central to many natural resource management problems, and yet there remains limited knowledge of how governance regimes change. An analytical framework is presented, which characterises governance regimes as multi-level learning processes. Within this framework, structural characteristics of governance regimes are identified as: (i) formal and informal institutions; (ii) state and non-state actors; (iii) multi-level interactions; and (iv) relative importance within bureaucratic hierarchies, markets and networks. Change is viewed as a social or societal learning process that moves step-wise from single to double, and triple loop learning. The authors present this conceptual framework as one way of emphasising a more systematic approach to analysing the structural characteristics of governance regimes and their influence on adaptive capacity and processes of dynamic social learning. Informal networks are considered to play critical roles in learning processes, although both formal and informal have relevance depending on goals and context. While governance regimes with greater complexity and diversity have enhanced adaptive capacity, single loop learning nonetheless tends to dominate adaptation responses. Greater sharing of relevant conceptual models and inter-disciplinary research approaches are needed in order to increase diagnostic capacity and to better inform policy in complex resource governance issues.