Surveillance of seat belt use in Kentucky is of vital importance in the effort to reduce trauma from motor vehicle crashes. Self-report health surveys provide an important source for epidemiologic data on risk factors associated with non-use of seat belts. However, estimates of the prevalence of seat belt use derived from such surveys are consistently higher than estimates obtained from observational studies. In the 1993 Kentucky Health Survey, we introduced a new question designed to narrow the gap between observed and self-reported seat belt prevalence. The new question, which asked what percent of the time seat belts are worn, was compared with a "traditional" question, which asked persons to describe their frequency of seat belt use as "always," "nearly always," "sometimes," "seldom," or "never." Of the 652 persons answering both questions, 65.4% stated that they "always or nearly always" used a seat belt -- a prevalence which was close to that reported by two other 1993 self-report studies: the Kentucky Behavioral Risk Factor Surveillance System and the Kentucky Health Interview and Examination Study. However, only 44.8% claimed to wear a belt "100%" of the time -- a prevalence much closer to the 41% of Kentuckians observed to wear seat belts in the same year. We recommend that health surveys adopt this new question and accept nothing less than "100%" as defining a seat belt user. We also recommend that physicians use the new "percentage" question when counseling patients on seat belt use.