Detection Response Task (DRT) Evaluation for Driver Distraction Measurement Application

Abstract

Research was conducted to support development of National Highway Traffic Safety Administration's (NHTSA’s) Phase 3 Driver Distraction Guidelines for auditory-vocal driver-vehicle interfaces. A single experiment was conducted in driving and non-driving test venues to evaluate the sensitivity of Detection Response Task (DRT) metrics to differences in attentional load. Three DRT variants were used: Head-mounted DRT (HDRT), Remote DRT (RDRT), and Tactile DRT (TDRT). A repeated-measures design required participants to complete all six combinations of DRT (3) and test venue (2). Secondary tasks included: 0-back, 1-back, and visual-manual radio tuning. Forty-eight participants provided two independent samples following the Phase 1 NHTSA Driver Distraction Guidelines selection criteria. Each 24-person sample had 6 participants (3 male, 3 female) in the following age ranges: 18-24, 25-39, 40-54, and 55+. Two identical stationary vehicles were used. One was connected to a fixed-base driving simulator where drivers maintained a constant following distance behind a lead vehicle. The other vehicle housed a non-driving venue in which participants performed DRT and secondary tasks without a concurrent driving task. Tasks were performed continuously for 3 minutes on each trial. Differences between test venues were more pronounced than differences between DRT variants. Response times were faster in the non-driving venue but differences between secondary task conditions were consistent across venues; radio tuning was associated with highest DRT performance degradation, followed by the 1-back and 0-back conditions, respectively. Non-driving hit rates were consistently higher than driving hit rates. A set of four planned comparisons and comparisons of effect sizes (ES) were used to evaluate metric sensitivity at different test durations. In the driving simulator, TDRT was slightly more sensitive than other DRT variants. In the non-driving venue, all three DRT variants provided comparable sensitivity for response time. Ceiling effects rendered hit rate data not sensitive to differences between conditions in the non-driving venue. A 2-minute data collection interval provided optimal sensitivity for testing. Based on (2-minute) small-sample (N = 24) testing in the driving simulator, both TDRT and HDRT detected all differences while RDRT performance was weaker. The TDRT had the highest level of test-retest reliability. For small sample non-driving venue testing, the 2-minute interval was also slightly better; however, there were no differences among DRT variants in sensitivity or test-retest reliability. Overall, the TDRT was consistently more sensitive than other DRT variants, albeit marginally, in the driving simulator. Using a non-driving venue requires reliance entirely on the response time metric. Differences between DRT variants were too small to identify a better performing DRT in the non-driving venue. Potential visual conflicts associated with RDRT and to a lesser extent with HDRT may create problems when used with visual-manual tasks in either venue.