Numerous published models can be used to predict curve speed based on geometric and operational characteristics like radius, superelevation rate, and approach tangent speed. Speed-based design consistency measures have also been developed to help identify which curves on a roadway are the most severe. However, the use of speed reduction alone can result in improper assessment of curve severity because drivers are more reluctant to reduce speed on roadways with higher speeds and thus accept speeds associated with higher crash risk. New measures of curve severity are suggested, based on considerations of side friction demand and kinetic energy. The increase in side friction demand above drivers' comfort thresholds is shown to be roughly proportional to the kinetic energy reduction associated with speed reduction. Agencies can use these curve severity measures to assist in identifying curves in their jurisdictions that would most likely benefit from safety improvements.