A study was conducted to investigate driver performance on curves. The between-trial factors were Blood Alcohol Content (BAC) level (0.00, 0.7, 0.12 %) and type of driving scenario (eventful versus uneventful). The within-session factors were edgeline width, type of curve-warning sign, and curve type. Twelve male drivers drove continuously for two hours on each of three nights. Each subject negotiated 150 curves during each two-hour drive. Curve radii ranged from 57.3 to 94.2 m (188 to 309.2 ft). Advisory speeds presented on curve-warning signs ranged from 32.2 to 72.4 km/h (20 to 45 mile/h). The driving simulator was a completely instrumented cab resting on a fixed base. The results showed that curve-entry speed increased as radius of curvature increased. Lateral position error was greater on the curve with the smallest radius and least on the curve with the shortest length. Heading error first increased then decreased as curve radius increased. Neither the amount of road used nor the mean computed lateral acceleration were related to curve radius, heading change or length. These results are attributed to the absence of lateral-acceleration cues in the driving simulator.