The pursuit of sustainability poses many common challenges to policymakers across the developed and developing world. At the same time there are processes encouraging environmental policy convergence and integration into other policy fields. This would suggest abundant opportunities for policy learning and transfer. However, experience has shown that wholesale borrowing of models from other policy fields, or across jurisdictions seldom takes place. Why is this?

Traditional 'rationalist' and linear models of policy see change in policy systems as discrete, well-defined events, frequently precipitated by a crisis due to a failure of the previous system, or by the discovery of a new environmental problem, such as the ozone hole. Policy is reformulated through a review process in which new scientific knowledge plays a major role. New objectives, responsibilities and routines are set, and the policy system settles down again to a period of normal business. In this account, most learning occurs for a short period during the process of policy formation. Outside these transitional moments, established principles are implemented through formal regulatory procedures. According to these models, policy learning does not occur freely because the opportunities for learning are few and far between.

Research conducted through the UK Economic and Social Research Council's Global Environmental Change Programme and elsewhere has enriched and substantially modified the rationalist account of policy processes. Not only are policy systems considerably more 'messy' - for instance, there is often no clear distinction between policy making and implementation, as Fineman's study of the UK Environment Agency makes clear - but learning and change are pervasive and continuous within policy systems. New knowledge, changing expectations and practical experience are being applied by policy actors at many different levels, in a process of 'adaptive social learning'. This suggests that there may be multiple opportunities for policy learning and transfer right across the policy system.

Yet learning runs into numerous obstacles and blockages. Environmental policy learning does not take place in a vacuum. It is influenced by many factors including the capabilities and beliefs of policy communities, the political and other resources available within policy networks, and by the availability of scientific and other knowledges. Knowledge is seen as a key ingredient of learning, and this may arise from a wide range of sources. Shackley, for example, shows how formal 'big science' played a crucial role in building a case for policy responses to climate change. In quite a different context, Hulme argues for the importance of informal local knowledges of farmers and herders in a study of community-based conservation approaches in Africa. Shifts in understanding therefore may arise from multiple sites, resulting in either more fundamental reframing of policy problems, sometimes challenging long-held conventional wisdoms, or more incremental changes focused on more marginal instrumental changes. The pace of policy change and the influence of new knowledge, as Stephens points out, is therefore highly dependent on context.

Whatever its source, new knowledge and the prospect of change that it brings, frequently threatens existing policy relationships and structures of power. Stephens' work on how urban environmental health research influenced public health policy in Accra and Sao Paulo, shows clearly how knowledge and power interact. New and existing knowledge are continually reconciled. Different meanings are attached to research results, depending on underlying beliefs about inequality and commitment to current policy objectives. In two studies concerned with implementation of environmental policy in the UK, Fineman and Jordan also argue that underlying commitment - or the lack of it - to policy objectives is crucial to the process of policy learning and institutional adaptation.

Where then does policy learning occur, and what are the routes for learning? One of the common findings of the studies reported here is that responses to scientific and practical knowledge are highly differentiated. Stephens identifies two processes which she names 'snowballs' (the accumulation of research impacts within policy elites) and 'whispers' (the reinterpretation of research findings in broader constituencies). Hulme contrasts the differing approaches to community conservation in Eastern and Southern Africa, and the limited transfer of ideas between the two regions. In some cases practical action precedes policy change, as in the case of community conservation initiatives in Namibia reported by Hulme. Such project activities may provide important 'witnessing' opportunities which, in the longer term, help shift the policy debate. A number of the studies reported here show how policy and implementation are deeply intertwined. As Fineman argues, 'street level bureaucrats' need to balance many contrasting objectives in a highly politicised social environment. Thus due to the discretion of implementors, particular policy outcomes may be very varied, even where a uniform regulatory and policy setting exists.

Another common finding is that in increasingly open and participative policy settings, at both the local and the global level, environmental policy learning is most effectively achieved by adopting a more flexible and iterative model of the policy process. In particular there is an emphasis on the need for convening processes in which the lessons from theory and practice can be mediated and appropriated by a range of key actors in the policymaking process. Shackley, for instance, argues that the great success of the scientific assessments within the Intergovernmental Panel on Climate Change (IPCC) in shaping climate policy is due to the interactive, two-way nature of the review process involving business, non-governmental organisations and government, together with scientists. The building of trust and common commitment among a network of participants in the policy debate is seen to be crucial. Stephens amplifies this by showing that 'ownership' of research findings was a crucial factor in their uptake by policymakers. Hulme too emphasises the need for experimentation in policy responses, and counsels caution in accepting easily transferable 'blueprints'.

So what are some of the lessons learned from these studies? A number of common themes can be identified:

• There is a need to accept change as pervasive in environmental policy, with a commitment to continuous and adaptive learning. Simple blueprint transfers of policies rarely work well.
• There is a need for user involvement in the research process, based on a model of interactive research and networking by researchers with policy actors.
• There is a need for inclusivity and flexibility in the policy process, with an emphasis on the building of trust among key stakeholders.
• There is a need for clear messages anchored to key policy problems which are identifiable by policymakers.
• There exist options to 'reframe' problems as a way of making them more 'legible' to policymakers.
• Policy change often takes time - especially if fundamental assumptions are challenged and political and other interests are affected.
• Contexts matter - political, bureaucratic and economic contexts influence the pace and form of policy change.