Abstract
The visual-manual distraction guidelines by National Highway Traffic Safety Administration (NHTSA) provide a method to assess potential distraction of secondary tasks. According to the method, one instance of “testable task” is performed once by 24 participants, and three glance measures (i.e., mean duration of off-road glances, total eyes off road time, and proportion of long glances) are used to decide whether the task imposes acceptable demand on the driver. The current study investigates variability from both participants and task instances (e.g., different telephone numbers used for dialing telephone number task) reflected in these glance measures, with repeated sampling of participants and repeated sampling of task instances. While NHTSA test method is vulnerable to both sources of variability, multi-level statistical modeling and sampling task instances offers a promising approach to overcome this limitation.