The present widespread interest in the use of electronic media for presents an unprecedented opportunity for leveraging the computational medium's strengths for learning. However, existing software tools provide only primitive support for online knowledge-building discourse. Further work is needed in supporting coordinated use of disciplinary representations, discourse representations, and knowledge representations. This paper introduces the concept of representational guidance for discourse along with results of an initial study of this phenomenon in face-to-face situations. The paper then considers the requirements for supporting asynchronous online knowledge-building discourse, finding existing computer mediated communication tools to be particularly deficient in supporting artifact-centered discourse. A solution is proposed that coordinates discourse representations with disciplinary and knowledge representations.