A computational model of mental processing speed in drivers who are holding a conversation suggests that having a passenger in the car is not a distraction, but it does make us more cautious and slower to make decisions.

A reality of living in much of the United States and many other parts of the world is that cars are still the primary means of transit. This is true even in cycle-friendly cities like Portland or Amsterdam, and temperate high-density cities like San Francisco or London where you could walk outside comfortably all year round.

Driving can be monotonous, and we are always looking for ways to liven it up – listening to catchy songs, playing 20 questions. Another way of making us feel better during our commutes, or reducing the monotony of long stretches of flat, straight highways in Nebraska, is to bring a friend or coworkers literally along for the ride. Not only does this put fewer cars on the road, but having someone to commiserate with in traffic is a definite plus. (Plus you get to use the carpool lanes.)

While more passengers in a vehicle is certainly a good way to reduce overall traffic, and probably a good way to make sure you take breaks and avoid highway hypnosis, having passengers in the car increases the amount of information to process – you need to attend both to the road and to any conversation taking place.

Surprisingly, talking to someone on a cell phone while driving, a topic that has come up several times, poses much greater risks than talking to an actual passenger in the car with you. This presents an interesting puzzle for research into distracted driving – why would a real-life conversation not pose the same threat to safety as a hands-free or handheld cell phone conversation? Two main accounts have been proposed: (1) people process less information from any one stream because driving and talking share cognitive resources; or (2) people may be more cautious to make driving decisions when confronted with more information inside their vehicle.

A recent article in the Psychonomic Society’s journal Attention, Perception, and Psychophysics sought answers to these questions. Researchers Gabriel Tillman, David Strayer, Ami Eidels, and Andrew Heathcote conducted a study looking at the impact of high cognitive load across three conditions: driving without distractions, driving and holding a conversation with a hands-free cell phone, and driving and holding a conversation with a passenger in what is known as the Detection Response Task (DRT).

The DRT is a standardized task used to evaluate cognitive load while driving. During this task, participants have to respond to a red light presented in their peripheral vision. The researchers also implemented a computational model of information processing in the diffusion modeling framework, which we will discuss in just a moment.

In the DRT task in this study, participants drove on a simulated freeway and either did not have a conversation, or engaged in a conversation with another participant, who was either present in the driving simulator with them, or was elsewhere and communicated over a hands-free cell phone. Participants were shown a red light in the periphery every 3 to 5 seconds and pressed a button on the steering wheel when they detected the light.

Participants were found to be substantially slower to respond to the light when they were engaged in conversation with another person, in both the hands-free cellphone case as well as when talking to a passenger. You can watch this video (with German subtitles) demonstrating someone in a driving simulator and wearing the DRT device that flashes intermittent red lights.

Despite the clear effects of conversation on reaction times, it is not clear why participants were slower to respond in the conversation conditions – was there an information overload, or were people more cautious when making judgments on the DRT task? Tillman and colleagues used a diffusion model to tease these two possibilities apart and to understand the consequences of conversation on the DRT.

Diffusion models are an increasingly common and flexible computational modeling framework that are especially useful for modeling decision processes, especially those resulting in reaction time data. We have discussed diffusion models before on this blog.

Diffusion models have been used to understand a variety of tasks and processes, from speech onset latencies, to judgments of brightness and cognitive control tasks. These models use a few different parameters to model behavioral data like the time it takes to make a decision. The models that the authors constructed varied in three parameters: drift rate, response thresholds, and non-decision times. Drift rate refers to how quickly participants are able to process that there is new information, and theoretically processing stimuli is slower when cognitive load is high because less information trickles in. Once enough information has been processed to get over the response threshold, participants initiate a response, which has some motor or component that contributes to speed in the form of non-decision time. By orthogonally varying experimental variables that selectively affect these three parameters, we can independently assess the role of each of the three factors that might explain slower reaction times in higher load conditions.

Decision model schematic from Fast-DM by Voss & Voss

The model results were unexpected. The only important parameter in predicting reaction times appears to be the location of the threshold; the results of testing different parameters in the model suggest that participants’ response thresholds change. That is, people become more cautious during challenging situations and wait longer for more information, even if there would have been enough information to make a response in a less cognitively demanding situation.

If the message you took away from this research is that talking to your friend or family member while driving still has negative consequences, you might be right. It is important to take these results in context – while it seems that talking to a friend in the car with you does not pose additional risks, it is certainly worse than driving without distractions, at least in the sense that it makes you slower to respond to a secondary task. So instead, the best take away message from this paper is the exciting intersection of computational models and important questions that impact the day-to-day lives of many people who get around by car.

Psychonomic Society journal article featured in this post:

Tillman, G., Strayer, D., Eidels, A., & Heathcote, A. (2017). Modeling cognitive load effects of conversation between a passenger and driver. Attention, Perception, & Psychophysics, 79, 1795-1803. doi:10.3758/s13414-017-1337-2