Two experiments were conducted to understand the processes through which 5th graders discriminate relevant from irrelevant information when solving mathematical story problems. Visual scanning was recorded and coded as directed toward relevant information, irrelevant information, the question, workspace, and elsewhere. Experiment 1 focused on the effects of irrelevant numerical information, and Experiment 2 focused on irrelevant qualitative information. The visual scanning results showed that higher achieving and lower achieving students generally used question-guided comparisons of the relevant and irrelevant information when they succeeded. A 2nd effective strategy was to discriminate during their initial reading of the problems. In addition, whereas higher achieving students flexibly varied their visual scanning to fit with problem difficulty, the lower achievers showed less flexibility.