We used a virtual-environment driving simulator to show that, as fog increases and therefore reduces the contrast of the driver's image, the apparent speed of the vehicle slows. Participants asked to drive at a certain speed drove faster as the scene became foggier.

An accurate representation of the current speed is normally provided to drivers by the speedometer. However, reading this instrument requires drivers to divert their gaze and attention from the road to the appropriate dial. In conditions of reduced visibility produced by fog, drivers are reluctant to divert their gaze from the road to the speedometer for fear of missing an object emerging from the fog2. Hence it is exactly in conditions of reduced visibility caused by fog that drivers rely on their own perceptual judgement of speed.

Thompson3 has reported that the perceived speed of a moving-grating pattern depends on its level of contrast. As contrast decreases, the grating appears to drift more slowly. This effect has been replicated several times (and extended to other stimulus patterns4), but others have failed to replicate it or have shown that perceived speed is unaffected by random variations in contrast5. So it remains an open question as to whether this illusion of reduced speed with decreasing image-contrast will occur in conditions akin to those encountered when driving in fog.

We tested for the perceived slowing of a visual scene by conducting two experiments in a virtual environment that simulated the view from a vehicle moving along a road. Such a set-up allowed us to manipulate the visual stimulus in a highly specific manner.

Our first experiment involved showing the observer two scenes. In one scene, the weather was ‘clear’; in the other, it was either ‘clear’, ‘misty’ or ‘foggy’. The physical speed at which the two scenes appeared to move at the same speed was calculated (Fig. 1a) using standard psychophysical techniques. Subjects perceived the foggier scenes to be moving more slowly6.

Figure 1: Sense of speed decreases in fog.
figure 1

a, Perceived matching speed averaged over all subjects (n = 5) in three weather conditions under passive viewing. Foggier scenes gave the impression of slower movement (one-way analysis of variance; P < 0.001). simulation was performed using a silicon graphics crimson reality engine that updated the viewing screen (60 Hz VDU monitor) at a rate of 15 Hz. ‘Fog’ was simulated by blending a partially transparent polygon over each pixel. For each pixel, the red, green and blue guns levels were recalculated (R′) from the original pixel values (R) drawn from memory according to the formula, R′ = 255F + (1 - F)R. ‘Clear’ conditions: F = 0, contrast (root mean square) = 0.61; ‘misty’: F = 0.5, contrast = 0.20; ‘foggy’: F = 0.8, contrast = 0.07. b, The actual speed driven under three weather conditions (n = 9). Target speeds: circles, 30 m.p.h. (48 km h−1); squares, 50 m.p.h. (80 km h−1); triangles, 70 m.p.h. (112 km h−1). Analysis of variance showed significant effects of target speed (P < 0.0001) and weather conditions (P < 0.0001), but no interaction (F < 1).

Does this perceptual change affect driving speed in a more realistic task? We trained subjects (in clear conditions) to ‘drive’ a simulated vehicle at set speeds along a winding road, using a brake, accelerator and steering column.

In the experimental phase, the subject was given a target speed and told to adjust the car speed (while steering along the road) to this target, using the accelerator and brake. Subjects accelerated and decelerated as they thought appropriate for as long as they pleased, and then signalled that they believed that they were travelling at the target speed. Trials in which the weather was ‘clear’, ‘misty’ and ‘foggy’ were randomly interleaved. As the scene became more foggy, subjects drove at faster speeds (Fig. 1b).

This finding suggests that the ‘blame’ for many such accidents may not lie solely in the irresponsible nature of the drivers but with an unfortunate quirk of our perceptual systems.