May 25, 2006
I didn't see you!
From http://hedgehog.lancs.ac.uk:8181/newsletters/may2003.pdf
Motorcycle accidents tend to involve other road users who often claim not to have seen them in time to avoid a collision. The traditional explanation for such accidents is that the motorcyclist is relatively inconspicuous compared with other road users. Therefore the way to reduce motorcycle accidents is to make motorcyclists more conspicuous. However, database studies and those accidents involving highly conspicuous police motorcycles tend to question if this hypothesis is the only explanation available.
The age of the ‘offending’ driver in motorcycle accidents does not have the customary peak associated with young inexperienced drivers. Experience appears not to teach the car driver much about motorcycles. Motorcyclists tend to have the right of way when a vehicle emerges into their path. Just because a motorcyclist is highly conspicuous this is no guarantee that another road user will see them. Evidence from fatal accident records implies that the use of conspicuity enhancers do significantly reduce the chances of a serious accident.
The Transport and Road Research Laboratory coined the phrase ‘looked but failed to see error’ referring to a set of circumstances where a driver accounts for an accident in the terms of failing to see. A prototypical form of such accidents involves collisions between motorcycles and cars. Motorcycle accidents tend to involve another road user who often claims not to have seen them in time to avert a collision. Research into the causes of motorcycle accidents has tended to make two basic assumptions: firstly, that the offending driver actually looks but then fails to see the motorcyclist; secondly, that this failure can be explained in terms of the relative lack of conspicuity of the motorcyclist.
This laboratory experiment suggests an alternative explanation for such accidents. Experienced and inexperienced drivers viewed video tape clips of approaching traffic at intersections. Subjects' eye movements were recorded in response to different search instructions. Under the conditions of this experiment, experienced drivers appear to use ‘pre-programmed’ search patterns directed towards areas of the road environment which are rich in information; there was little evidence of these in the eye-movements of inexperienced drivers. Experienced drivers appeared to start their search at a midpoint in the scene whist inexperienced drivers started their search nearby. One consequence of this was that experienced drivers took longer to detect motorcyclists who were nearby. (i.e. to the left of the initial point of fixation.)
Drivers’ visual search at junctions can be very rapid. Models of human object recognition, such as Biederman's Recognition By Components (RBC) theory, provide a theoretical account for how we can rapidly comprehend a scene, despite short display times. Can such theories account for failures in driver's search at junctions? Inspired by RBC theory, this laboratory study attempts to understand changes in search behaviour by drivers at differing levels of experience. Each subject watched video clips of an approaching vehicle in a traffic scene, as viewed from the perspective of a driver at an intersection. In one of the clips, the vehicle was replaced by an animated shape. This was either a vertical bar moving in the same way as a motorcycle; a horizontal bar moving in the same way as a car; or a pixelated “rippling” distortion. Drivers’ decision times were recorded in response to differing search instructions: they were asked to search either for a motorcycle or for any motor vehicle.
Experienced drivers - but not non-drivers or inexperienced drivers - treated the animated shapes as vehicles, provided the search instructions matched the shapes orientation. Thus if the shape was a vertical bar and the instructions were to search for a motorcycle, experienced drivers were more likely to believe that a vehicle was present than if they had been asked to search for any vehicle.
An unexpected finding was that police traffic officers treated the “rippling” distortion as a cue to a vehicle’s presence; possibly movement is all that is required by such highly trained drivers. We suggest that with experience, drivers may develop shorter search times at junctions and may extract from complex traffic scenes only a minimal amount of information, based on prior expectancies about what they are likely to see.
The moral is quite apparent. If you drive a car, look harder. If you ride a bike, assume that the driver will look but might not see!