Less-than-expected fatality reductions after seat-belt legislation has been introduced in a jurisdiction may be explained in terms of selective recruitment of parts of the driving population and/or behavioral adaptation by beginning belt users. The present investigation has compared the relative merits of these two hypotheses at the level of individual driver behavior. In the initial study the driving behavior of groups of habitual wearers and nonwearers of the belt was compared. Nonwearers made two trips, one with the belt on and one without the belt. Habitual wearers drove belted only. The main part of the experiment was a 105 km freeway route. Two additional tasks of a somewhat more critical nature, a double lane-change manoeuvre and the performance of a braking manoeuvre in front of a fixed obstacle, were performed after the freeway trips. Factor analysis on 39 variables describing driving behavior on the road and during the additional tasks resulted in five factors. One of these, the factor describing the distribution of driving speed on the freeway, differentiated between nonwearers and wearers (thus yielding support for the selective recruitment hypothesis) as well as between wearing and not wearing the belt by the same drivers (thus yielding support for the behavioral adaptation hypothesis). In the follow-up study the original wearers and nonwearers were assigned to one of four experimental treatments: (i) the promise by the experimenter of a considerable incentive for not having a culpable motor vehicle accident over a period of a year. Half the habitual wearer subjects were assigned to this condition. The expectation was that this group would become more careful in their driving; (ii) a control group, consisting of the remaining habitual wearers; (iii) the agreement between the experimenter and the subject that the latter would buckle up in everyday driving for the year to come--half the habitual nonwearer subjects were assigned to this condition; (iv) a control group to the previous treatment, consisting of the remaining habitual nonwearers. All subjects returned for repeat measurements, consisting of the freeway trips plus additional tasks, three times over the next year. The main result was that beginning wearers (group iii) showed signs of continuing behavioral adaptation, in the form of increased speed and increased propensity for close following, as well as several minor effects. The "incentive" group (group i), however, did not change driving behavior in the expected sense, i.e. in the safe direction.