The overarching research problem for this study was the need to improve upon rational models of policy analysis and delivery, to suit complex postmodern implementation environments. A theoretical model suited to implementing and evaluating major education reform initiatives was devised. Called the 'Multiple Discourse Model', it was grounded in systems theory, containing elements reminiscent of social systems, organisational and structural functionalist research, especially that of Hoy and Miskel (1982)1. However the model was also designed to incorporate a parallel naturalistic analysis reminiscent of postmodern critical pragmatic approaches, such as those explored by Cherryholmes (1994)2. Over a period of five years, this model was developed through an evaluation of the implementation of the National Aboriginal and Torres Strait Islander Education Policy (AEP) in the Australian Capital Territory government secondary schools sector. The distinguishing feature of the study's methodology was its multiperspective analysis, an approach suggested by Mclaughlin (1987)3 to take account of the differing communities of discourse which exist in a reformist policy implementation environment. To operationalise the research problem, dimensions of policy effectiveness were articulated. These were addressed through a comprehensive set of research indicators, extracted from the AEP's national policy goals and the local strategic and operational plans. Data aimed at judging the effectiveness of implementation were collected from multiple sources using multiple research instruments. These data were analysed in three stages using a purpose-designed computer program which could cross-reference between the four interacting dimensions of research indicators, research instruments, data sources, and potential variables modifying policy/program outcomes. It was found that this model produced clear conclusions about the effectiveness of AEP implementation in the delimited sector, within the framework of the AEP's own policy assumptions. The model also provided insights into critical issues which are generalisable to the national context, such as the power of cultural hegemony and the socio-political predicament of Aboriginal and Torres Strait Islander dispossession. As a methodology, the model was found to have a number of technical advantages, including its capacity for focussing on selected areas of the implementation environment, its provision of access to multiple levels of detail amongst data and its possession of mechanisms for monitoring its own internal validity. The evaluation case study, used as the vehicle for the Multiple Discourse Model's development, demonstrated that best-practice administration was in place which enhanced short and medium-term policy/program outcomes. However, the study's findings also suggested that a fundamental disjuncture existed between the AEP's policy/administration paradigm and the conflicting assumptions of the primary target communities, reinforcing the findings of Sykes (1986)4. The research results suggested that despite measurable successful inputs, the planned long-term outcomes of the AEP will not necessarily be achieved. No significant administrative structures or actions were apparent which could resolve this lack of synchrony at the interface between government delivery systems and 'grass roots' Aboriginal and Torres Strait Islander community values. This raised doubts about whether any systems approach, however well refined, could be socially useful not only for evaluation, but also as a basis for reform policy and public administration in a postmodern pluralist democratic setting. The evaluation was therefore used as a locus for theoretical reflection as well. A new policy paradigm is suggested, based on a power-sharing 'theory of community', more in keeping with Aboriginal and Torres Strait Islander peoples' aspirations for self determination and more likely to alleviate the so far unresolved destructive effects of cultural and political dispossession.