The Impact of N-back-Induced Cognitive Load and Time Budget on Takeover Performance

Cognitive load refers to the quantity of mental resources that are demanded from a human during information processing within a given task. Despite cognitive load being highly implicated in driver distraction, the effect it has on takeover performance remains elusive. Some studies find no effects of cognitive load on the timing and quality of takeovers, whilst others have found potential increases in crash risk and delays in response times. The effect of time budget – the time between the onset of an event and an impending crash - is much clearer; drivers tend to react faster when time budgets are smaller. However, no study has investigated whether the effects of a pure cognitive load interact with the effects of time budget during critical takeovers from a hands-off Level 2 automated system. The current study uses a Bayesian multilevel modelling approach to quantify the direction, size, and uncertainty of the effects that cognitive load and time budget have on driver performance. Drivers (N = 37) participated in two experimental drives using a hands-off Level 2 automated system. During one drive, participants completed a pure cognitive load task (2-back); during the other they only monitored the road. After approximately 2 minutes of automated driving, a lead vehicle decelerated with a short (TTC = 3 s) or a long (TTC = 5 s) time budget, generating a potential rear-end conflict. There was no impact of 2-back-induced cognitive load on the timing or quality of takeovers. Drivers had similar braking and hands-on steering wheel reaction times, and there were no consistent or substantial effects of cognitive load during post-takeover manual driving. Conversely, time budget was strongly associated with takeover performance. Drivers were faster to respond but more erratic in their post-takeover lateral control following events with shorter TTCs. We discuss the reasons for an absence of an effect from the N-back-induced cognitive load on takeover performance; this is despite subjective and gaze-based evidence indicating that drivers were cognitively loaded. One suggestion is that rear-end scenarios are inappropriate for understanding the detrimental effects on takeover performance imposed by a cognitive loading task. Alternatively, we suggest that scenarios which incorporate decision making processes (such as lane change manoeuvres or hazard perception tasks) might be better for elucidating the effects of cognitive load during transitions of control.