Consensus conferences evolved as a response to the public's increasing dissatisfaction with technocratic decision-making processes that are judged to have repeatedly failed to serve its interests. The staging of the first Australian consensus conference at Old Parliament House in Canberra in March 1999 therefore presented an ideal opportunity to analyse the evolution of this new kind of policy input from its conception through to its implementation and subsequent evaluation. This thesis set out to provide an analysis of that trajectory using elements of the theoretical approach known as actor-network theory (ANT). Previous analyses of consensus conferences have generally provided only limited evaluations of single aspects of the entire process of setting up, implementing and evaluating such a conference. Furthermore, many of the early evaluations were conducted by reviewers or units which were themselves internal to the consensus conference under scrutiny. My own analysis has tried to offer broader, although inevitably less detailed, coverage, using a perspective from contemporary social theory that offers particular advantages in analysing the creation of short-term networks designed for specific purposes. By describing and analysing the role of this relatively new policy-making instrument, I have explored the different sub-networks that operate within the consensus conference process by focussing on the ways in which the conference was organised and how the relationships between the organisers and the participants helped to shape the outcomes. Thus the entire consensus conference sequence from idea to outcome can be thought of as a construction of a network to achieve at least one immediate goal. That goal was a single potential policy input, a consensus position embodied in the report of the lay panel. To realise that goal, the network needed to be recruited and stabilised and its members made to converge on that collective statement. But how is it that a range of disparate actors, including lay and expert, are mobilised to achieve that particular goal and what are the stabilisation devices which enable, or fail to enable this goal to be reached? In the context of the first Australian consensus conference, three key alignment devices emerged: texts, money and people. Yet it is clear from the evidence that some of these network stabilisation devices functioned poorly or not at all. This thesis has drawn attention to the areas in which they were weak and what importance that weakness had for the kind of policy outcome the consensus conference achieved. The role and extent of these powerful stabilisation devices in networks was therefore a vital issue for analysis. If one of the criteria to evaluate the success of a consensus conference is that it provides the stimulus to hold another, then the Australian conference must be deemed so far a failure. No further Australian consensus conference is planned. However, Australia stands to forfeit a number of advantages if no further consensus conferences or similar occasions are organised. Policy formation in contemporary democracies has had to accommodate an increasing array of new participants in order to track more effectively the diversity of potentially significant opinions on complex policy issues. This process requires new and transparent ways to educate and inform the public on policy issues and to ensure that policy makers are better informed about the needs and concerns of their community. As the evidence presented in thesis for the Australian example and its predecessors overseas suggests, consensus conferences have the potential to play a role in the contemporary policy-making context. But the realisation of that potential will vary according to their institutional contexts and the capacity of the actors to create the temporarily most stable and productive network out of the heterogeneous human and material resources to hand.