This paper introduces the concept of representational guidance for collaborative learning discourse and summarizes the results of two studies: a laboratory study investigating differences in discourse process and a classroom study investigating differences in products of the investigations. The laboratory study compared three representational tools for recording and reasoning about evidence during collaborative inquiry. Sixty students from introductory secondary school science classes participated. The tools were unconstrained text (a word processor), graphical evidence mapping, and a matrix or tabular representation. Participants were provided with hypertext-based materials pertinent to a problem in which multiple explanations were proposed for a disease. Analysis of discourse transcripts was consistent with predictions concerning the amount of talk about evidential relations. The classroom study focused on the effects of representational tools and of assessment rubrics on learning to evaluate diverse sources of empirical evidence and coordinate these data with domain theories. Four high school science classes participated. The representational tools were graphical evidence maps and text. The assessment rubrics provided specific criteria for scientific inquiry, encouraging self-reflection and peer coaching. The results indicate that evidence mapping provides better support than prose, as evidenced by amount of information and quality of inferences recorded in classroom artifacts. The assessment rubrics improved the quality of work of students using the evidence mapping tool but not those using prose. (Contains 26 references.) (Author/SLD)