Students are frequently presented with novel visualizations introducing scientific concepts and processes normally unobservable to the naked eye. Despite being unfamiliar, students are expected to understand and employ the visualizations to solve problems. Domain experts exhibit more competency than novices when using complex visualizations, but less is known about how and when learners develop representational fluency. This project examined students' moment-by-moment adoption patterns for scientific visualizations. In a laboratory experiment, introductory-level organic chemistry students viewed familiar ball-and-stick and novel electrostatic potential map representations while solving chemistry problems. Eye movement patterns, verbal explanations, and individual difference analyses showed that students initially relied on familiar representations, particularly for difficult questions. However, as the task unfolded, students with more prior knowledge began relying upon the novel visualizations. These results indicate adoption and fluent use of visualizations is not given; rather, it is a function of prior knowledge and unfolding experience with presented content. (Contains 4 figures.)