Evaluation of the Total Eyes-off-Road Time Glance Criterion in the NHTSA Visual-Manual Guidelines

Abstract

NHTSA issued a 12-s total eyes-off-road time (TEORT) glance criterion for its visual-manual guidelines simulator test. This criterion relied on NHTSA’s analysis of manual radio-tuning data from its simulator study using a 2010 Toyota Prius premium radio and its test-track study of five 2005 to 2010 vehicles using various radio-tuning control types. However, this 12-s criterion was biased falsely downward. First, NHTSA oversampled the youngest participants (with the shortest TEORT scores) in the simulator study and oversampled the newest radios (with the shortest TEORT scores) in the test track study. Second, NHTSA incorrectly assumed that track and simulator metrics had a one-to-one correspondence, but the track TEORT was substantially shorter than the simulator TEORT for a matched Prius task with matched older participant ages. Third, NHTSA set its criterion at exactly the 85th percentile of the TEORT data rather than its upper confidence limit; hence, traditional radio-tuning tasks would fail repeated tests simply because of random test-to-test variation. The current study, after adjusting for these and other biases, estimated a criterion of 21.4 s from the identical NHTSA data sets. Emerging knowledge that relies on naturalistic driving studies suggests that cumulative glance metrics such as TEORT have low predictive validity for crashes compared with other metrics. A new test should be developed with metrics of higher predictive validity for relative crash risk on the basis of a growing scientific consensus about the relationship—both positive and negative—between eye glances, relative crash risk, and secondary task performance while driving.