This study employed an eye tracking technology to explore university students’ visual attention and learning performance while learning Japanese using an interactive manga-based e-book. The developed e-book consisted of 8 pages accompanied by 13 annotations with both text and graphical formats. The subjects consisted of 60 students whose eye movements were tracked and recorded by the eye tracking system. These students came from the applied foreign language department in a northern university in Taiwan, of which 30 were assigned to high prior knowledge (PK) group and the other 30 were assigned to low PK group. Eye tracking measurements, including total contact time, number of fixations, latency of first fixation, and number of clicks on the defined regions of interest of the two groups were compared to indicate their visual attention. The results revealed that overall students spent more time on reading text and annotation than graphic information. The high PK students showed longer fixation durations on the texts, while the low PK students showed longer fixation durations on the graphics and annotations. Meanwhile, the low PK students used more clicks to look up underlined annotations whenever they didn’t know words or phrases on the e-book. In addition, with respect to the latency of the first fixation, the graphic captured the attention faster than the text because of the size and its appeal to the students. Further analysis of saccade paths indicated that the low PK students showed more inter-scanning transitions not only between the text dialog and the annotation zone but also within annotation zone. Finally, the results of reading comprehension pretest and posttest found that there was a significant difference in learning outcomes between each PK group.