This study proposed an analytic framework for coding students' dialogic argumentation and investigated the characteristics of the small-group argumentation pattern observed in modeling-based learning. The participants were 122 second grade high school students in South Korea divided into an experimental and a comparison group. Modeling-based learning was applied to the experimental group in a topic of apparent motion of Mars on the other hand, explanation-based classes led by a teacher were provided to the comparison group with the same topic. Students created a paper model of Mars retrograde motion and learned through small-group discussion about the astronomical phenomena which is represented by the created model and the reason that caused these phenomena. The analytic framework for coding students'argumentation in the modeling-based learning was composed into eight components and three categories including model-related statement, claim and reasoning. The results showed that a variety of model exploration is essential for useful argumentation which includes the reasoning added with the claims. By investigating the time sequential pattern of the argumentation, a process of argumentation frequently cross all three categories was revealed as an effective pattern in modeling-based learning. Additionally, the results of comparing the concepts of apparent motion of Mars in the experimental group and comparison group, the ratio of students with correct concepts of the experimental group was higher than those of the comparison group. The implication of these finding for modeling-based learning environment,productive modeling discourse are discussed.