OBJECTIVE: This study investigates the interaction between human perception and driver behavior on horizontal curves, focusing on how road geometry and visibility affect driving performance. METHODS: A driving simulator replicated 3 curve types by radius-200 m (sharp), 400 m (moderate), and 600 m (loose)-under day and night conditions. The focus of expansion (FOE) is the source point of optical flow, and an FOE model was established to linked the driver's visual perception with vehicle dynamics. Data on eye movement and vehicle dynamics were collected from 24 drivers (mean age: 27 years, mean driving experience: 3.8 years). RESULTS: The results indicate that driving at night on sharp curves significantly impairs the ability of drivers to align their perception with vehicle motion, leading to delayed steering adjustments and increased lateral errors. The most dangerous areas, identified as the back half of the test curves and corresponding to the minimum FOE radius, were where the misalignment between perception and motion was most significant. On loose curves, decreased driver vigilance was observed, potentially due to a perceived reduction in steering demands, underscoring the role of psychological and contextual factors during curve negotiation. CONCLUSION: This study underscores the importance of optimizing curve radii and enhancing the alignment between drivers' visual perceptions and vehicle dynamics to reduce accident risks. In real traffic, placing traffic guidance mid-curve may better prompt drivers to slow down, particularly at night. Integrating FOE-based feedback into advanced driver assistance systems (ADAS) could further enhance performance by offering real-time cues tailored to curve geometry in low visibility.