This paper demonstrates how a driver’s performance in a single-vehicle road departure avoidance maneuver may be characterized using mathematical expressions. Nine single-valued covariates are needed to fulfill the expressions:​ five to describe the pre-maneuver vehicle state (speed, heading angle, edge distance, road curvature, and curve entry distance) and four to describe the driver response (brake application time, deceleration level, steering input time, and steering angle). A procedure to find the best set of covariate values is demonstrated using a trial from a series of test track experiments in which subjects maneuvered along a Jersey barrier. The procedure provides a high level of conformity between the actual vehicle path with respect to the barrier and the path derived from the nine covariates. Thus, the entire avoidance maneuver may be faithfully described by a set of mathematical expressions. Subsequently, each of the thousand-plus test track trials is characterized by a single, ninecovariate data record (instead of several time-histories made up of thousands of records, one for each time point). Ultimately, such a reduction in data benefits the development of an in-vehicle crash warning system. By structuring the avoidance maneuvers in a recordlevel dataset, warning system alternatives may be investigated directly by applying traditional statistical analyses on a large collection of records. The paper discusses extensions of the method to the theoretical possibility of continuously estimating the nine parameters in real time as part of a collision avoidance driver assistance system.