Excessive speed is one of the primary safety hazards facing highway users. Too often, operating speeds exceed design speeds, and speed limit postings are generally ignored by drivers. There is renewed interest in designing roadways that encourage drivers to operate at or below design speeds. There is not, however, a reliable body of concomitant knowledge regarding what combinations of design properties should be used to accomplish this goal. To date, only limited interrelationships have been explored, but these explorations have not begun to approach evaluations of the complexity of the design environment, which encompasses combinations of design properties such as lane width, shoulder width, clear zone, horizontal and vertical curvatures, and even the type and intensity of vegetation along the corridor. This research integrates the use of virtual reality simulations of various highway sample designs. These sessions can simultaneously accommodate significant numbers of subjects and gather their feedback through electronic audience response systems. The response feedback is modeled through the use of a fuzzy set system that allows designers to inspect the interrelationships of many different design parameters. Such modeling systems are particularly well suited to nonlinear, data-sparse applications such as this. This research reports on a pilot data modeling effort, in which the discomfort responses of 15 participants to 22 different virtual roadway design combinations were gathered in less than 30 min. These data were sufficient for use for modeling the way in which more than 400 possible roadway design combinations would affect the participants' perceptions of speed and safety.