AASHTO's Policy on Geometric Design of Highways and Streets contains geometric design criteria for horizontal curves for new and major reconstruction. These criteria are based on design side friction factors that were established in the 1940s. The objective of this research was to establish a probabilistic approach to the design of horizontal curves and compare the results with current design criteria. Reliability analysis commonly is used in structural engineering and recently has been used in stopping sight distance research for transportation design. In this research, design variables were considered random instead of deterministic to incorporate variations that occurred in the field (e.g., driver heterogeneity or differences in tire performance characteristics) into the analysis. The effects of wet pavements and tire characteristics for passenger cars and heavy trucks were considered. The superelevation rate was considered as a design input, and minimum radii for a target reliability index were presented for combinations of input mean speeds and superelevation rates. All other design inputs were considered random variables (e.g., available pavement friction), along with their appropriate distributions, in the reliability analysis. The results show that a target reliability index of 3.0 is appropriate for use in the probabilistic design of horizontal curves. Furthermore, when only skidding failure is considered, the design criteria for passenger cars can reasonably accommodate heavy trucks. The recommendations for future research consider probabilistic methods and more complex models of horizontal curve forces.