A total of 83 drivers, 51 males and 32 females, aged 25-65, were recruited to drive an apparently unmodified passenger car for 1 h over at least four laps of a predetermined route on public roads, which included seven sets of traffic lights and sections limited to 50, 70, 90 and 110 km/h. They were randomly assigned to one of two thermal conditions (21 or 27 degrees C), and drove only during the hours of daylight. A computer initiated unprepared signals to which drivers would normally be alert. Drivers responded by pressing a foot-switch and reporting verbally. Signals were selected at random from 21 possible signals, and were presented for up to 3 min, with a random delay of 30-180 s after each response or failure to respond. The negative effect of heat stress on vigilance was statistically significant. At 27 degrees C, the overall proportion of missed signals was 50% higher and response times were 22% longer than they were at 21 degrees C. These effects of heat were significant and proportionally greater in the second half-hour, for subjects < 40 years and for speeds below 60 km/h (i.e. in city traffic). The latter finding suggests that heat may have increased arousal, and there was some indication of a redistribution of attention away from the most peripheral signals at the higher temperature. Overt driving errors were observed significantly more often at 27 degrees C than at 21 degrees C for women only.