In science education, it is vital for teachers to consider students' academic and emotional needs. Teachers' prediction of students' learning states has been commonly regarded as an indicator to measure that competence to understand students. This study aimed to explore the outcome and the process of prediction to reflect teachers' pedagogical content knowledge. The study was to detect the differences in preservice teachers' eye movement behaviors with eye-tracking technology between the prediction group and nonprediction group. The prediction group predicted the option students will most likely choose for a given question, while the nonprediction group solved the problems on their own. The result showed that preservice teachers in the prediction group were more considerate of students' ideas and review the information among different areas of interest when they were required to detect the problem from the perspective of students. In the prediction group, preservice teachers with positive prediction focused more on the correct option repeatedly, while those with negative prediction tended to inspect carefully within each incorrect option. In addition, successful and unsuccessful problem solvers in the nonprediction group responded to problems in a different manner, in which successful problem solvers paid more attention to inspecting information from options, including both correct option and incorrect options.