Understanding the ways that unconscious information processing occurs can help educators teach in more efficient ways. In any given science task, there are typically a myriad of relevant and irrelevant features that frame and direct reasoning pathways. To better train students to navigate task features productively, more needs to be understood about the ways that attention-attracting task features interact with reasoning processes. Eye-tracking methodologies allow physics education researchers to record where a student allocates their attention, including the initial focus of attention. In this paper, we discuss results from eye-tracking research into a mechanic graph task that has a strong intuitive reasoning pathway that leads to an incorrect conclusion. We analyze data from students in algebra-based physics courses. We find that analysis of the eye-tracking metrics of first look and dwell time are consistent with predictions made from dual-process theories of reasoning.