The purpose of this study was to explore how students with high- and low-prior-knowledge employed multiple representations in argumentation evaluation and generation tasks. The argumentation performance and eye-movement behaviors of 96 college students in these tasks were investigated. The number of participants who proposed complex argumentation levels and the argumentation accuracy was higher in the high-prior-knowledge group than in the low-prior-knowledge group. Moreover, the high-prior-knowledge group demonstrated greater eye-movement transitions between representations compared with the low-prior-knowledge group. Both groups had greater transitions in the generation task than the evaluation tasks. The high-prior-knowledge group distributed attention to representations with more flexibility, revealing that they were more aware of the task requirements and more able to employ multiple representations for arguments. In the argumentation evaluation tasks, the high-prior-knowledge group performed referencing behaviors in the reading sequences between representation text and equation and between representation table and figure, whereas the low-prior-knowledge group was inclined to look back and forth between representation text and table. In the argumentation generation task, the two groups displayed similar reading sequences. It indicated that learners with higher knowledge may perceive the similarity between homogeneous representations and constrained interpretations of the complex representations by using easier representations, or further integrated representations to achieve deeper understanding, which then improved their argumentation performance. Implications for improving learners' argumentation skills in a multirepresentational display are discussed.