The purpose of this mixed-methods study was to explore the effect of varied gender groupings on argumentation skills among middle school students in Taiwan and the United States in a project-based learning environment that incorporated a graph-oriented computer-assisted application (GOCAA). A total of 43 students comprised the treatment condition and were engaged in the collaborative argumentation process in same-gender groupings. Of these 43 students, 20 were located in the US and 23 were located in Taiwan. A total of 40 students comprised the control condition and were engaged in the collaborative argumentation process in mixed-gender groupings. Of these 40 students, 19 were in the US and 21 were in Taiwan. In each country, verbal collaborative argumentation was recorded and the students' post essays were collected. Among females in Taiwan, one-way analysis of variance (ANOVA) indicated that statistically a significant gender-grouping effect was evident on the total argumentation skills outcome, while multivariate analysis of variance (MANOVA) indicated no significant gender-grouping effect on the combined set of skill outcomes. Among females in the US, MANOVA indicated statistically significant gender-grouping effect on the combined set of argumentation skills outcomes Specifically, U.S. female students in mixed-gender groupings (the control condition) significantly outperformed female students in single-gender groupings (the treatment condition) in the counterargument and rebuttal skills. No significant group differences were observed among males. A qualitative analysis was conducted to examine how the graph-oriented computer-assisted application supported students' development of argumentation skills in different gender groupings in both countries. In each country, all teams in both conditions demonstrated a similar pattern of collaborative argumentation with the exception of three female teams in the US. Female teams, male teams, (the treatment condition) and mixed-gender teams (the control condition) demonstrated metacognition regulation skills in different degrees and with different scaffolding.