Abstract: As a policy process that generates and disseminates policy knowledge,policy trial experiments have drawn extensive attention from scholars in many fields,however,the related research has failed to address how policy knowledge is constructed during the process of from pilot trials to official implementation(congdian daomian "point-to-surface"). This paper examines a science and technology policy case study by analyzing discourse networks to reveal how the experiences of a single trial could become "best practices" in the process of discourse dissemination. We notice that once the pilot experiment is told,it enters a complex discursive construction process. First,there is a dynamic construction process of the pilot experiment by discourse variation and intertextuality. On the one hand,pilot trial units,mass media,and academic experts engage in interpreting the experiment according to their own editing rules. On the other hand,through interactions among various interpretations over the period of expanding policy implementation,a discourse consensus on policy knowledge would gradually emerge,which is then adopted as official policy discourse. Second,the study explores the impact of different editing rules on the construction and diffusion of policy knowledge,the discursive order and rules that make up the interpretive framework of policy knowledge,as well as the complexity of social practices behind the process of intertextuality. Lastly,the study offers a new perspective explaining why sometimes policy experimentation fails to be widely adopted. When the discursive interpretation becomes so essential to the construction of pilot experiences,it may lead to the actual lessons being lost in translation. By introducing a new research angle of discourse,this study provides some new insights on the studies of policy experimentation and policy knowledge.

Key words: policy experimentation, policy knowledge, constructivism, translation, discourse network analysis