Research Papers

Conspicuity of motorcycles in traffic: Evidence from two change-blindness experiments

Version 1
Date added June 18, 2015
Downloaded 0 times/fois
Category 2015 CARSP XXV Ottawa
Tags Research and Evaluation, Session 1A
Author/Auteur Bertrand Sager, Elisabeth Kreykenbohm, Thomas M. Spalek
Stream/Volet Research and Evaluation

Slidedeck Presentation Only (no paper submitted)

1A - Sager

Abstract

BACKGROUND
Motorcyclists are often involved in collisions where the driver of a car “looked but failed to see” the motorcycle before turning left across the motorcyclist’s path. Previous research mostly focused on perceptual factors that contribute to motorcycle inconspicuity, such as size, lighting, and colour. This research often suggests countermeasures aimed at increasing motorcycle visibility, such as fluorescent jackets, headlight modulators, and daytime running lights. However, despite motorcyclists’ efforts to increase their conspicuity, collision statistics remain unchanged. The assumption that motorcycles are “invisible” in traffic remains largely unchallenged. Here, we employ a change-blindness paradigm to test the assumption that motorcycles are less visible than cars in a dynamic traffic environment. This work follows our previous change blindness experiments which used static images of traffic scenes.

AIMS
This study aims to test the hypothesis that motorcycles are less conspicuous than cars in dynamic traffic environments. To this effect, we employ a change blindness paradigm in a driving simulator. Targets (cars or motorcycles) are removed from the scene during a screen flicker, and participants report whether they detected a change. If motorcycles are less conspicuous than cars, we expect that the detection rate for changes involving motorcycles will be lower than the detection rate for changes involving cars.

METHODS
In a first experiment, participants drove a high-fidelity driving simulator down a straight road populated with an equal number of cars and motorcycles randomly parked on both sides of the road. Every 400 metres, the screen flickered and, on half of the trials, one of the vehicles (either a car or a motorcycle) was removed from the scene. After each flicker, participants reported whether they noticed a change or not. Signal-detection analysis was used to determine whether there were any differences in either sensitivity or bias between cars and motorcycles.
In a second experiment, participants drove towards a busy intersection and on half of the trials a vehicle was removed from the ambient traffic during a screen flicker. On each trial, participants indicated whether they detected a change, and signal-detection measures (d’ and β) were again collected and analyzed for the effect of location (target is on the left, right, or center), possible conflict (target is entering or exiting the intersection), vehicle type (target is a car or motorcycle) and lane position (target is in a near or far lane).

RESULTS
Data collection is ongoing for Experiment 2, but preliminary results indicate that motorcycles are detected as frequently as cars in busy traffic environments. The results of Experiment 1 show that motorcycles are detected slightly more frequently (60.46% of the time) than cars (56.7% of the time), t(58) = 2.03, p = .047.

DISCUSSION
The assumption that motorcycles are difficult to see, or inconspicuous, in traffic is primarily based on the statements of drivers that collide with them and report a failure to see the motorcycle, a report that seems credible in light of motorcycles’ relatively small size. Consequently, countermeasures aimed at decreasing motorcycle collisions have focused on making the motorcycle more visible, generally by applying various lighting treatments. Unchanging collision statistic, however, suggest that the problem may not be conspicuity related. Here, we replicate and extend previous work where we found that motorcycle, car, and pedestrian detection rates were identical in static driving scenes, this time in a more ecologically valid paradigm.

CONCLUSIONS
(Contingent on the results of Experiment 2)
Motorcycles are detected at least as frequently as cars in the two dynamic change-blindness experiments presented here. We suggest that right-of-way violation collisions are not due to low motorcycle conspicuity, but rather to oncoming drivers’ inability to judge when it is safe to turn. This account is consistent with the finding that car-motorcycle right-of-way violation collisions differ from car-car right of way violation collisions only when the motorcycle is proceeding straight through an intersection and when the car driver turns left across its path (de Craen, Doumen, & van Norden, 2014), and that a motorcyclist’s lane position affects an oncoming driver’s ability to judge whether it is safe to turn because of the paucity of motion-cues offered by the motorcycle when it is approaching directly along the line of sight (Sager, Yanko, Spalek, Froc, Bernstein, & Dastur, 2014).

Bertrand Sager, Elisabeth Kreykenbohm, Thomas M. Spalek