Research Papers

Visual Attention Failures during Turns at Intersections: An On-road Study

Filename 1B-Kaya-FP.pdf
Filesize 412 KB
Version 1
Date added July 9, 2018
Downloaded 11 times/fois
Category 2018 CARSP XXVIII Victoria
Tags Research and Evaluation, Session 1B
Author/Auteur Kaya, Ayas, Ponnambalam, Donmez
Stream/Volet Research and Evaluation

Slidedeck Presentation 

1B -Kaya


Crash data indicate that misallocation of attention is a major source of vehicle crashes with vulnerable road users (pedestrians and cyclists) at intersections. Video recordings from outside and inside the vehicle indicate that drivers allocate their attention based on their expectations but the extent that drivers fail to scan for vulnerable road users at intersections is not known. In this paper, we examine failures to check for vulnerable road users during right turns at intersections. Eye-tracking data was analyzed from 19 drivers between the ages of 35 and 54 who participated in an on-road instrumented vehicle study conducted in downtown Toronto. Each participant made two right turns from a major arterial road. In addition to attention allocation failures, we assessed whether the objective data was correlated with experience driving in the area as well as with drivers’ subjective responses about their intersection-related errors collected through the Driver Behaviour Questionnaire (DBQ). Eleven of the 19 participants had a failure in at least one of the intersections; all failures related to checking for cyclists. At a marginally significant level, attentional failures were more likely for those who drove more frequently in downtown Toronto and for those who had larger error scores on intersection-related questions of DBQ. The prevalence of attentional failures observed is alarming, especially given that our participants represented the lowest crash-risk age group. It appeared that drivers less familiar with an area were more cautious when it comes to negotiating an intersection. Additionally, drivers appeared to be aware of their intersection-related errors as indicated by their DBQ responses. Further research with an increased sample size and on a variety of intersections is needed to generalize these findings.