One feature of design-based research is progressive refinement of classroom practice and learning environments through the cycles of design and analysis. In our previous study (Yamamoto et al., 2013), we taught a science unit to fifth graders that dealt with the construction of arguments comprising claim, evidence, and reasoning. Through an analysis of the students’ performance after the unit, in this study, classroom practice was revised and its effectiveness was tested. The revisions to the unit were aimed at fostering the transfer of acquired argument skills in line with two design principles: detailed reflection on written arguments and fading scaffolds. One hundred and fourteen fifth graders were taught the revised unit, and then performed two argument tasks. The first argument task required students to write an argument related to the content of the unit after the training. The second argument task, the transfer task, required them to write an argument on other scientific topics which they had learned before. Their performance on these tasks was compared with that of students in the previous study. While students who participated in the revised unit obtained the same scores at the first argument task as students in the baseline unit, they showed more improvement in the transfer task in criteria concerning evidence. More students could justify their claims with both evidence (data) and reasoning (known scientific rationale). These results showed that the two design principles were effective in fostering the transfer of argument skills through the iterant cycles of design and analysis.