Cognitive Distraction in the Wild

Strayer et al. (2015) have put together a series of experiments that are the first to give clear evidence that increases in cognitive workload lead to decrements in indices of safe driving performance, not only in the laboratory on a driving simulator but in the field with an instrumented vehicle. The study achieves another first by showing that both driver responses signaled by attention-grabbing cues (e.g., looming and brake light onset) and driver responses signaled by top-down cues (e.g., glance behaviors) are compromised by cognitively loading secondary tasks. Thus, critically for the field, the authors’ paper cements what has been suspected for some time: that increases in cognitive distraction lead to decreases in the measures that index safe driving performance.

If there were uncontroversial evidence that the increasing levels of distraction were associated with corresponding increases in crashes, the research would be useful but not altogether surprising, in my opinion. However, given the fact that some recent studies have shown just the opposite, at least when cell phone conversations are the distracting activity (Klauer et al., 2014), there is much to ponder, especially about the effects of cognitive distraction in the wild on crashes. I believe that the authors’ research points to some of the reasons that there may be a discrepancy and possible next steps one might undertake to determine the complex relation that may exist between cognitive distraction in the wild and crashes. I will speak about two.

First, the relation between cognitive workload, both in the lab and with instrumented vehicles, and driver performance could be the same as it is in naturalistic driving. However, the control or baseline condition in the lab and the instrumented vehicle may not be representative of the baseline level of attention given to driving in the field. This is not entirely speculation, being based on a recent study on the role of mind wandering in crashes (Galéra et al., 2012). Briefly, in that study, over 1,000 drivers involved in automobile crashes who were brought to an emergency room were asked no later than 5 hr after the crash about the driver’s thoughts just before the crash. Among those drivers interviewed, fully 52% said that their minds were wandering before the crash and 17% said that they were extremely distracted.

The question that this study raises in the current context is whether the average level of cognitive distraction caused by the cell phone is less than the average level of cognitive distraction caused by mind wandering. If that were the case, then cell phones could have a prophylactic impact when their effect is measured on crash rates in the field (and the appropriate control condition involves considerable mind wandering) but have just the opposite impact when their effect is measured in the lab or with instrumented vehicles (and the control condition is one that involves a driver paying relatively complete attention to the forward roadway).

Second, there may be a net advantage to the decrease in the side-to-side glancing that occurs during cognitively distracting activities that does not occur during driving without such activities (a decrease noted in the authors’ study as well as many other studies). Certainly many crashes occur because the driver has his or her eyes off the forward roadway. Studies suggest that both experienced and novice drivers take many more especially long and therefore dangerous glances to the side of the road than they do inside the vehicle (Divekar, Pradhan, Masserang, Pollatsek, & Fisher, 2013) and are thereby more likely to miss a vehicle ahead that is stopping. Drivers on the cell phone are more likely to have their eyes on the forward roadway. Thus, there may be a net benefit to cell phone use over no cell phone use.

In summary, in my opinion, there are some obvious next steps to take, given the strength of the authors’ study. Cognitive distractions do lead to decreases in indices of how safely drivers are behaving when those measures are taken on a driving simulator and with an instrumented vehicle and the control condition is the fully attentive driver. But cognitive distractions in the wild (on the open road) may have a prophylactic benefit, a benefit that can be studied quite independently of the resolution of whether there is a real benefit on the open road. The authors’ index of cognitive distraction and their use of event-related potential and eye movements point us in the direction of those next steps.

First, with respect to mind wandering, one would want to run an experiment whereby the route was a familiar one without the embedded Detection Response Task (DRT), perhaps repeating the same hour-long drive over the course of 10 sessions. Such an experiment is more apt to capture mind wandering than is the typical half-hour experiment (total driving time) because the driver has become familiar with the route, the route is a long one, and the DRT does not interfere with mind wandering. Presumably, physiological measures would best index the extent of mind wandering. The question here is whether the driver in the 10th session is engaged in much more mind wandering than is a driver who is specifically asked throughout the drive to pay as complete attention as possible to the driving task.

Second, with respect to scanning, one would again want to run the same experiment, only this time using eye movements as the dependent variable. Over time, one would want to gather information on the proportion of glances to the side when the driver both is and is not cognitively loaded. The information in Table 1 indicates how a driver 3 times as likely to crash when loaded—p(crash|front) = 0.0003—is, overall, less likely to crash than a driver who is not loaded simply because the driver who is not loaded is spending considerably more time glancing to the side—p(side) = 0.2. Note that the relative risk of crashing, expressed as the ratio of p(crash|no load) to p(crash|load), is equal to 0.61, exactly what is obtained in the naturalistic studies for novice drivers.

OPEN IN VIEWER

Table 1:

Relative Risk of Crashing When Cognitively Loaded and Not Loaded

	No Load	Load
p(crash|side)	.01	.01
p(side)	.2	.1
p(crash|front)	.0001	.0003
p(front)	.8	.9
	p(crash|no load)	p(crash|load)
	.00208	.00127

With respect to the authors’ article, my analysis suggests that any measure of cognitive distraction that is going to be tied to crashes needs to include the effect of a particular cognitively distracting activity on the prevalence of other behaviors that might be still more dangerous as well as the relative risk of crashing when the driver is and is not engaged in the other behaviors.