|
Abstract
|
In much of the world, fear of traffic crashes limits healthy and sustainable behaviors such as walking and biking. A possible root cause is based in perception: Drivers may sometimes fail to see pedestrians or bicyclists, increasing the probability that they will be crashed into. Here, we developed a method for directly measuring detection of vulnerable road users (VRUs: pedestrians, bicyclists, or motorcyclists). The results reveal a bias in perception that may be critical for understanding driver behavior.
Methods. Observers detected VRUs from brief glances of largish street-scene pictures, presented (250 ms) left or right of fixation. The scenes were crowded (one or more autos present) or uncrowded (no autos). Crowding was expected to reduce accuracy of identification. Observers responded "yes" or "no" regarding the presence of a VRU, and received training with representative stimuli before testing. VRUs were present in 50% of the scenes, in appropriate (legal) locations.
Results and Conclusions. Overall accuracy (percentage correct) was near 75%, and higher for uncrowded scenes than for crowded scenes. Critically, signal detection analyses indicated that the errors were not distributed equally; there was a bias for observers to miss VRUs, and a corresponding low false alarm rate. The bias occurred in both crowding conditions, but was stronger with crowded streets. The bias was consistent across observers (24 of 27). Further, the bias resulted in an especially high miss rate for pedestrians in crowded scenes -- 65% of pedestrians were missed. The results indicate that fears of VRUs are well-founded. Drivers fail to detect VRUs, assuming that they are not present, especially when traffic is more congested. This may be an instance of the error of attention (Chabris & Simons, 2009) -- "if I didn't see it, it wasn't there." This perceptual error can result in deaths in VRU crashes.
Meeting abstract presented at VSS 2013
Language: en |