Visual search of traffic scenes

This study investigates top-down governed visual selection in natural traffic scenes. It is assumed that optimal scanning of the visual environment takes place by using appropriate search routines. In the present experiment subjects had to search for a target object (i.e. traffic sign, other road users) which was embedded in natural traffic scenes. Given a particular prototypical scene, the target was located either at a likely (expected) or unlikely (unexpected) position. The results show the existence of scene dependent scanning strategies: if the target object is at an expected location, search is somewhat faster than when it is located at an unexpected condition. Search in the unexpected condition was significantly more error prone than search in the expected condition. This suggests that subjects strategically prepare for the upcoming stimulus and only search those locations which are likely to contain a target. If the target is not found at those likely locations, subjects tend to give a negative response. The effect remained after repeated presentation. The importance of these findings for search during actual driving is discussed.