The dissemination of successful educational innovations has long been a strongly supported policy in educational settings. Recent research developments have led to a debate as to the appropriateness of a centralized versus decentralized dissemination policy. The current research provided a test of the modified Research, Development and Diffusion (RD&D) model of innovation dissemination. Results clearly support the notion that a modified RD&D model is indeed feasible as a means of bringing about large scale educational change. Well specified programs were adopted and implemented within conservation developer-defined acceptable bounds of fidelity. The significance of the fidelity of implementation was addressed through its relationship with the degree to which the model program was reinvented at the local adopting site, the extent to which the program has become routinized at the local site and the effectiveness of the implementation at the local site. Policy implications were discussed. (Author)