Introduction

Driving a car can be a cognitively demanding task in home-work-home commuting scenarios since road hazards are an undercurrent concern. Several studies examined the effect of cognitive demands on driving performance, especially in terms of safety implications (cf., Elfering et al., 2013; Turgeman-Lupo & Biron, 2017). However, surprisingly little is known about the role of cognitive load on daily drive in terms of (i) the effect of the demands of driving on subsequent cognitive tasks carried out just after arrival at work, and (ii) the effect of working cognitive demands on the driving home.

Cognitive load (or mental workload) refers to the association between the demands of a situation or task and the mental resources required to perform it (Brookhuis & de Waard, 2010; da Silva, 2014). Driving can be a highly variable task involving irregular navigation patterns, and the ability to override automated behaviors to promote an effective mapping of motor-sensory inputs, internal states, and outputs (Miller & Cohen, 2001; Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977). Since continuous monitoring and high levels of alertness are essential to driving (Cao & Liu, 2013), the levels of cognitive load required can vary greatly. Indeed, driving relies on the ability of drivers to keep a car within the boundaries of a road lane, at a safe distance from the car ahead, and at a reliable speed in relation to the surrounding traffic, as well as to attend to unpredictable stimuli like the appearance of pedestrians, changes in traffic lights, and irregularities on the surface of roads, to cite just a few examples.

Consistent with the cognitive control hypothesis (Engström et al., 2017), high levels of cognitive load occur when a new and demanding driving task strains the cognitive capabilities of the driver (Warm et al., 2008), while driving under low levels of cognitive load regularly relies on the underlying automated behaviors of experienced drivers. Increases in cognitive load linked to executive functions like the planning and preparation of future actions (Lorist et al., 2000; Shanmugaratnam et al., 2010; van der Linden et al., 2003) and the control of behavioral inhibition can have a disruptive impact on driving performance (Platten et al., 2014; Tabibi et al., 2015).

Variations in cognitive demands caused by workload can modulate driving behaviour, and daily driving can significantly impact the mental health of workers (Amponsah-Tawiah et al., 2016; Novaco et al., 1990; Novaco & Gonzalez, 2009; Rowden et al., 2011), including general fatigue and assorted forms of distraction (cf., da Silva, 2014; Lee, 2008; van der Linden et al., 2003). Stretching cognitive effort across multiple tasks can impair cognitive performance (Howard et al., 2020; Innes et al., 2021), but the effects of distracted driving (i.e., distraction hangover, persistence of distraction) are difficult to assess because the specification of what constitutes distraction during driving is complex (Hurts et al., 2011; Snider et al., 2021), and observed both when cognitive demands were either too high or too low (Lal & Craig, 2001).

While prolonged periods of high cognitive load can increase mental fatigue and impair performance (Murphy et al., 2016), low cognitive load can lead to a decrease in attention levels, which in turn can impair information processing accuracy (Warm et al., 2008). Notably, impaired performance due to decreases in attention levels are a significant cause of road accidents (Borghini et al., 2014; Grandjean, 1979; Mizuno et al., 2011). In parallel with such findings, another study reported that driving performance under moderate levels of cognitive load had a beneficial effect on performance (Pavlidis et al., 2016), showing that not all activities that increased cognitive processing during driving necessarily compromised driving safety (Hickman & Hanowski, 2012).

It is important to keep in mind, however, that in previous studies the effect of cognitive load was examined during rather than before driving. And, to the best of our knowledge, the effect of driving (with its own cognitive load) on subsequent cognitive tasks has not been directly investigated despite a growing number of studies on cognitive carryover and readiness for action effects (cf., Miller, 2018; Vercillo et al., 2018). Hence, this study examined the relationship between distinct but contiguous levels of cognitive loads in two scenarios of daily commuting (home-to-work and work-to-home) to address unanswered and relevant questions:

  1. (i)

    Does a moderately demanding period of driving affect the cognitive performance of a task carried out soon afterward? We predicted that a higher level of cognitive load would have an adverse impact on accuracy in comparison to a lower level of cognitive load (control) condition.

  2. (ii)

    In turn, does a set of demanding cognitive tasks affect the performance of a subsequent driving task? We predicted it would affect driving speed, ability to stay within the confines of a lane, and distance kept from the leading car on a virtual motorway.

The findings are discussed in terms of the activation of attentional networks triggered by tasks requiring moderate levels of cognitive load, which could spill over into other timely contiguous tasks, with potential implications for work-related performance and driving safety in scenarios of daily commuting.

Study 1. Effect of driving on subsequent cognitive performance

Some of the core cognitive demands during driving are associated with executive functions (e.g., attention to the road, impulse control, journey planning) (cf., Vilchez et al., 2023). Here we examined if driving—with its own cognitive load— would impair or benefit the performance of a post-driving cognitive task.

A favorable review by the Faculty Research Ethics Committee was received and the study complied with the recommendations of the British Psychological Society and the tenets of the Declaration of Helsinki.

Method

Participants

There were 104 participants, but the data from six participants were excluded for assorted reasons (outlier values for reaction time, accuracy, or age). From the remaining 98 participants, 40 participants completed the first condition of the study (35 females, 5 males; Mage = 26.65 years, SD = 8.33, age range = 19–46), and 58 completed its second condition (51 females, 7 males; Mage = 24.59 years, SD = 5.66, age range = 19–45); some were full-time undergraduate students (~ 70%), and others were members of the public. They were recruited through internal faculty emails, social media, or word-of-mouth. Data collection was carried out at the start of the second semester, a period usually free from exams (in students’ case).

All participants held valid driver’s license for at least six months (M = 6.72 years, SD = 5.73). Participants reported normal neurological functioning, normal or corrected-to-normal visual acuity, and were fluent in English. All participants could opt to be entered into a prize draw, and Psychology students could opt to receive research participation credits instead.

As there were no previous studies addressing the research questions raised here, the sample size was loosely based on studies on cognitive load during driving, and it relied on opportunity sampling for two four-months periods over two years.

Measures and tasks

Multi-Source Interference Task (MSIT)

Cognitive interference occurs when the processing of one stimulus feature impedes the simultaneous processing of a second stimulus attribute (Bush et al., 2003). The MSIT in this study was a monochromatic stimulus-response positional conflict task inspired by well-known interference tasks (e.g., Stroop, Eriksen flanker, Simon effect). It contains multiple dimensions of cognitive interference affecting executive functions such as attention, impulse control, novelty, target detection, and error detection. The task was not language-specific, and the written instructions were easy to understand and retain, even though it was demanding in terms target detection, visual attention, impulse control, etc.

Each MSIT trial had three single digits presented next to each other; two of the digits were repeated (distractors) and one was unique (target). Participants had to identify which number was the target and press the correspondent number on the keypad. For example, in the trial stimulus “1 1 2”, the target is “2” and the participant had to press “2” on the keyboard as fast as possible. Participants were given 10 practice trials and instructed to answer as accurately and as quickly as possible. Once they had confirmed they understood the task, participants completed further 200 trials/each. Feedback was provided (correct vs. incorrect) and MSIT was run using E-Prime® 2.0 software.

Simulated driving task

Participants sat on an XPDS fixed-base driving simulator (model Silverstone SG01) with software designed by the same company, XPI Simulation (Fig. 1). The driving “motorway follow 1” scenario was presented in a forward view with a 3-screen configuration. The middle screen included a small panel to represent the rear mirror view of a car. Smaller panels were inset on the left and right-hand screens to simulate wing mirror views. The tasks in the seat-based driving simulator were used to induce a moderate cognitive load in the participants.

Fig. 1
figure 1

The XPI driving simulator (left) and a schematic representation of the virtual driving path (right), with the distance between the ego and the lead car (“stock headway”). Dotted lines define the lanes

Full size image

All participants had a training session in the driving simulator before the experimental session. Driving performance was sampled every 0.12 s across the driving session and automatically averaged, resulting in approximately 4880 measures/participant.

Procedure

After providing their written consent, participants were allocated to one of the two driving conditions: Low Cognitive Load (LCL) or Moderate Cognitive Load (MCL).

The virtual driving scenario required participants to follow a lead car and drive fast enough to keep close to it, but at a safe distance. The lead car had a constant speed (ca. 70 mph; 105–110 km/h) and the drivers had to follow it into different lanes avoiding other cars on the virtual road. Several lane changes were required during the 10 min drive, but the exact number of lane changes and cars on the other lanes per minute was not provided by the software developers.

While driving, participants also had to complete a verbal task. An audio file with sequences of three single digits (e.g., 2 2 1) was played and, after each sequence, participants had to say aloud which was the position of the unique number in the sequence. In example 2 2 1, the number ‘1’ was unique and it was in the third position, so the correct answer was ‘3’ or ‘third’. Once the driver had replied aloud, the next sequence was played, until the driving session was completed. Performance in the verbal task during the driving session was not recorded, but a researcher was present during the session to make sure the drivers had provided an answer to all digit sequences presented.

Participants in the MCL condition completed the driving task and then moved to the table with the laptop to complete the MSIT, which was ready to be run. Participants in the LCL condition did not take part in the driving simulation session; instead, they were asked to “wait until the setup for the cognitive MSIT task was completed” (i.e., up to 10 min).

The between-subjects design was used; performance was evaluated with cognitive tests novel to all participants, but the higher cognitive load (MCL) condition was induced through driving, while the lower cognitive load (LCL) condition did not include it.

Results

Speed, stock headway, and stock lane difference

In the MCL condition, the mean driving speed was 107.62 (± 4.26 S.D.) km/h, the mean stock lane difference was 0.53 m (± 0.69 S.D.), and the averaged stock headway was 14.67 m (± 6.06 S.D.). As mentioned before, there was no driving task in the LCL condition.

MSIT accuracy

The MSIT accuracy in the LCL condition was 78% but it raised to 88% in the MCL condition. There was strong evidence that the difference in accuracy between the two conditions was significant, t(96) = 3.44, p < .001 (Fig. 2a).

Fig. 2
figure 2

Accuracy (a) and reaction time (RT) (b) in the Multi-Source Interference Task (MSIT) in the low cognitive load condition (LCL: no-driving) and the moderate cognitive load condition (MCL: demanding driving). The boxplots on the right of each of the plots show the averaged accuracy and RT, and the 95% confidence intervals for the two cohorts

Full size image

MSIT reaction time

The reaction time in the LCL condition (740.94 ms) was faster than the reaction time in the MCL condition (755.18 ms), t(96) = − 0.36, p = .030, pointing to a speed-accuracy trade-off (Fig. 2b).

Study 2. Effects of cognitive load on subsequent drive

The findings in Study 1 suggested that a moderate cognitive load (through driving) had a significant and positive effect on the performance of another timely contiguous cognitive task. This study examined if the reverse was true, i.e., would a task with a moderate cognitive load affect a subsequent performance on a driving task?

As before, a favorable review by the Faculty Research Ethics Committee was received and the study complied with the recommendations of the British Psychological Society and the tenets of the Declaration of Helsinki.

Method

Participants

From the 36 participants who completed the first day of testing, 31 returned to complete the second day of testing (23 females, 8 males; Mage = 24.87 years, SD = 6.43, age range = 19–40 years); 23 participants were students, 13 were members of the public. As in Study 1, the data were collected at the start of the second semester, a period usually free from exams. All participants held a valid driver’s license for at least one year (M = 5.61 years, SD = 6.65), had normal or corrected-to-normal vision, and were fluent in English. As in the first study, participants could opt to be entered into a prize draw, and Psychology students could opt to receive participation course credits instead.

Since there were no previous studies addressing the research questions raised here, the sample size was loosely based on studies on cognitive load during driving, and it relied on opportunity sampling for two four-months periods over two years.

Measures and tasks

Mental Fatigue Questionnaire (Chalder et al., 1993)

Mental fatigue was assessed for screening purposes only. A Likert scale (0–3) was used to assess difficulties in concentrating, thinking clearly, and memory impairments (e.g., “Do you feel sleepy or drowsy?”, “Do you have difficulty concentrating?”, “Do you think as clearly as usual?”).

Cognitive tasks

The Cambridge Neuropsychological Test Automated Battery (CANTAB; Cambridge Cognition Ltd, UK) was used to induce a moderate cognitive load (MCL) and low cognitive load (LCL). The tests were chosen based on cognitive skills required in driving or working: impulse control, attention, and general alertness. Pilot trials showed that participants had no difficulties understanding and completing those tests and the performances were in the normative cognitive range for non-clinical samples.

The cognitive task in the MCL condition consisted of three tests requiring a considerable level of cognitive processing: Affective Go/No-Go, Stop Signal Task, and Rapid Visual Information Processing Test. The combination of tests lasted approximately 30 min (20–40 min range).

  • Affective Go/No-Go (AGN): used to assess the extent to which impulsivity depends on a participant’s emotional state (affective impulse control) by measuring processing biases for negative and positive stimuli. There are 10 blocks of words, each block comprising 28 words presented briefly in the centre of the screen (duration = 300 ms, ISI = 900 ms). Participants must press a response button only when a positive or negative target word appears on the screen.

  • Stop Signal Task (SST): used to measure response inhibition (impulse control). There are five blocks of 64 trials. On each trial, an arrow appears within a circle pointing either to the left or the right. Participants must press a corresponding left or right key depending on the direction of the displayed arrow, unless a ‘beep’ noise occurs, in which case no key should be pressed. The outcome measures used were the proportion of successful stop responses and mean RT.

  • Rapid Visual Information Processing Test (RVIP): used as a measure of sustained attention. Digits from 2 to 9 appear in a box in the centre of a screen at a rate of 100/minute. Participants must identify specific sequences of 3 digits (3-5-7, 2-4-6, or 4-6-8) appearing within the stream of numbers by pressing the right push button when they identify the target numeric sequences.

The cognitive task in the LCL condition had two tests: Simple Reaction Time and Graded Naming Test. The combination of tests lasted between 5 and 10 min.

  • Simple Reaction Time (SRT): used to test general alertness and motor speed. Participants must press a button whenever a white square appears on the centre of the screen.

  • Graded Naming Test (GNT): used to assess object-naming ability and detect impaired language functioning. Participants are asked to name aloud the objects or animals depicted in black-and-white line drawings shown on the screen.

Simulated driving task

The “motorway follow 1” driving scenario was the same one used in Study 1, but with no added verbal task. Drivers had to follow the car into different lanes avoiding other cars on the virtual road (Fig. 1). The driving performance of the licensed drivers in the simulator was assessed in terms of speed, stock headway, and stock lane difference. As before, driving performance was sampled every 0.12 s across the driving session and automatically averaged, resulting in approximately 4880 measures/participant.

  • Stock headway: it refers to the measurement (in meters) of the forward-backward distance deviation between the front bumper of the simulator’s (Ego) car in relation to the back bumper of the Lead car being followed, indicating the driver’s ability to keep a safe distance from the target (~ two cars).

  • Stock lane difference (or lane keeping performance): it refers to the lateral deviation of the simulator’s car (in meters) in relation to the lane lines used by the virtual car been followed. It indicates the driver’s ability to track and maintain lane position. Lateral deviations increase the likelihood of accident, as it leads to drifting towards other road users or obstacles.

Procedure

Participants provided written informed consent and demographic information before starting the study. A practice driving session with the seat-based simulator was carried out before the data collection started to allow participants to familiarize themselves with the driving simulator with the pre-programmed driving scenario. Before the start of the study, participants were also asked to answer the mental fatigue questions.

The cognitive tests were presented on a laptop placed on a table in the same room as the driving simulator; a button box attached to the screen by a cable was used to record the responses. The order of the two conditions of cognitive load was pseudo-randomized and counterbalanced and had to be completed in two consecutive days. Participants were randomly allocated to the MCL or LCL condition on Day 1. On Day 2 they completed the remaining condition (i.e., everyone completed the LCL and the MCL conditions). The research assistants monitored the completion of the cognitive tasks before the participants were asked to move to the driving simulator, but feedback on the tasks was not provided.

After completing the cognitive tasks, participants were asked to drive for 10 min in the “motorway follow 1” pre-programmed driving scenario and to keep a constant speed and distance from the car ahead, as well as stay in the same lane as the lead car. A stopwatch was used to control the 10 min driving time. If a participant drove off the road or was slow at the point of stopping (ca. 5% of trials), the stopwatch was paused, and once the participant was ready to recommence driving, the stopwatch was restarted until the 10 min driving was completed. The data from the simulator was automatically stored for each driving sequence and then collated for data analysis.

As the driving performance can be affected by one’s driving experience (which was not the same for all participants), a within-subjects design was used to test the same participant in the two experimental conditions.

Data analysis

The speed, stock headway, and stock lane difference mean values were extracted for each of the 31 participants who completed the MCL and the LCL conditions. A GLM univariate repeated-measures was carried out with each of the variable scores and cognitive load condition (LCL, MCL) as a within-subjects factor. Post hoc Bonferroni tests were also conducted.

Results

Self-reported levels of mental fatigue were tested at the start of each of the two days of testing and no significant differences were observed, t(30) = 0.35, p = .729.

Speed and stock lane difference

There was no significant difference in speed between the LCL (105.60 km/h) and the MCL (106.44 km/h) conditions, F(1,30) = 1.79, p = .191, η² = 0.056, and no significant difference in speed SD, F(1,30) = 0.004, p = .950, η² = 0.002. Equally, there was no significant interaction between speed and the run order of the LCL and MCL conditions (i.e., condition order); F(1,29) = 0.31, p = .583, η² = 0.01.

The means for the stock lane difference in the two conditions (LCL = 0.63 m; MCL = 0.63 m) were similar, F(1,30) = 0.06, p = .814, as it was the case with the response variability (i.e., stock lane difference SD), F(1,30) = 0.19, p = .666. The analysis was repeated with the condition order as a between-subjects factor, but again no significant interaction with stock lane difference was observed, F(1,29) = 0.29, p = .596, η² = 0.010.

Stock headway (distance from lead car)

All drivers kept a relatively safe distance from the stock car (i.e., car in front of them). The safe distance between cars in the LCL condition was reliably shorter (-7.19 m) than that in the MCL condition (-11.54 m), F(1,30) = 5.45, p = .026, η² = 0.154, Cohen’s d = 0.419. Those responses had a similar degree of variability, F(1,30) = 3.28, p = .080, η² = 0.099, and the order in which the two conditions were run did not affect the findings, F(1,29) = 0.12, p = .727, η² = 0.004 (Fig. 3).

Fig. 3
figure 3

Relative distance kept from the lead car on the driving simulator. Cognitive load before the driving session: moderate (MCL) and low (LCL).  Positive values indicate that the driver’s car was relatively closer to the lead car than the recommended safe distance, while negative values indicate that the driver’s car was further away from the lead car than the recommended distance. Each dot represents a “driver”. The inset shows the means for the two conditions and the bars represent the SE

Full size image

Discussion

The possible effect of a cognitively demanding driving task on a separate but timely contiguous task had not been examined in previous studies, and neither had the effect of a demanding cognitive task on a timely contiguous driving task. That gap was addressed in this study.

In Study 1, a driving task with a moderate level of cognitive load (MCL) was carried out before participants were asked to complete a subsequent cognitive task, the Multi-Source Interference Task (MSIT). A no-driving condition was used to mimic a lower cognitive load (LCL) condition. The experimental setup was used as a proxy for a home-to-work commuting scenario. It was hypothesized that a demanding driving period would negatively affect the performance on the subsequent MSIT when compared to the performance after the LCL condition. It could be argued that the improved accuracy in MSIT was due to practice with impulse control through the verbal task performed during driving, which also required impulse control. Although such speculation cannot be entirely dismissed, practice effects may have been minimal because the verbal task carried out during driving did not involve the visual attention or motor coordination needed to perform well in MSIT.

The findings partially refuted our hypothesis and showed a speed-accuracy trade-off: MSIT accuracy in the MCL (driving) condition was higher than the accuracy in the LCL (no driving) condition, but MSIT reaction time was longer under a higher than lower cognitive load. Such findings are not peculiar since earlier studies showed that both low and high levels of cognitive load can have detrimental effects on driving. Low levels of cognitive load common to “easy” driving situations are typically linked to boredom, while high levels of cognitive load are often linked to driving in complex environments (e.g., new surroundings, heavy traffic) which require heightened multisensory integration, spatial processing, and visuospatial attention to process salient information relevant to safe navigation (Broadbent et al., 2023; Naert et al., 2018; Wahn & König, 2017). A study examined eye-tracking behaviour during driving and showed that attending to distracting tasks increased cognitive load by loading working memory capacity (Broadbent et al., 2023). On the other hand, intermediate levels of cognitive load can be beneficial, which can lead to a performance resembling an inverted-U curve effect similar to the Yerkes-Dodson Law and to psychological phenomena like the Mere Exposure Effect.

In Study 2, the experimental scenario simulated a work-to-home commuting by car scenario. Cognitive tests from the CANTAB suite were used to induce moderate levels of cognitive loads short before participants were asked to carry out a subsequent driving task. The findings showed that a higher cognitive load before driving had no significant effect on driving speed or lane keeping ability, but participants kept a longer distance from the lead car in comparison to the condition with a lower cognitive load.

The findings were aligned with our initial hypothesis and with the posited adverse effects of cognitive load by the Resource Theory (Hurts et al., 2011; Wickens, 2002, 2008), even though an increased distance from the lead car could be seen as a positive effect in terms of driving safety rather than an impairment in performance. Indeed, drivers talking on a cell phone also tended to increase their following distance from the lead car compared to controls (Strayer et al., 2006). The increased distance kept from the lead car on a virtual motorway associated with a higher cognitive load could point to some level of awareness about an impaired cognitive processing, which in turn could have prompted a compensatory increase in alertness leading to a more cautious, safer driving behaviour. Interestingly, drivers reduced their speed in response to prolonged periods of demanding cognitive load (Brookhuis & de Waard, 2010; Curry et al., 1975; Mizuno et al., 2011), but in this study no change in speed was observed in the MCL condition.

Limitations and future studies

It is worth noting that in previous studies all tasks were concurrent activities rather than temporally separated, which precludes direct comparisons with the findings from this study. Furthermore, despite the importance of cerebellar contribution to the coordination of movement, its relationship with executive functioning and cognitive load is poorly understood (Koziol et al., 2014) making it hard to predict if it is involved in the effects reported here.

The setup of experimental paradigms with tasks triggering distinct levels of cognitive load simulating commuting scenarios was complex. One could argue that instead of using the two conditions LCL (no driving) vs. MCL (driving plus verbal task) in Study 1, one could have used LCL (driving without verbal task) vs. MCL (driving plus verbal task). However, LCL and MCL conditions had to be sufficiently different to test the subsequent cognitive performance efficiently. Afterall, the sole presence vs. absence of the verbal task during driving might not have been sufficient to create two distinct levels of cognitive load. Equally, it is possible that the cognitive tasks used in the MCL condition in Study 2 cannot be compared with the LCL condition because they had different durations.

Another limitation is related to the definition of what constitutes a “moderate cognitive load”. Even though the MCL conditions in the two studies are likely to have had an intermediate level of cognitive load (in comparison to other studies and control conditions), a precise quantification of such load is complex and beyond the scope of this study. For example, it has been suggested that at least two hours of an activity such as driving is needed to generate an increase in cognitive load large enough to induce strong levels of mental fatigue likely to compromise executive processing (cf., Brookhuis & de Waard, 2010; Mizuno et al., 2011).

Largely, the findings presented here tag along with Anderson’s Activation Theory, which posits that once a brain network is activated, its nodes spread the information relevant to the focus of attention facilitating the processing of the task ahead, as observed in semantic memory studies (Anderson & Pirolli, 1984; Heise & Westermann, 1989; Huff et al., 2021). Indeed, the activation of supervisory attentional neural networks triggered by cognitive load is known to facilitate adaptable and non-routine behaviors (Norman & Shallice, 1986) core to driving. Several studies investigated the propagation of neural activation (Bartlett, 1943; Deco et al., 2015; Desmond & Matthews, 1997; Engström et al., 2017; Ridderinkhof, 2014; Shine et al., 2016; Vercillo et al., 2018; Victor et al., 2015), which does not always cease immediately after a task stopped (McCormick et al., 2022; Salo et al., 2017; Totoki et al., 2010). For example, Strayer and colleagues (2022) showed that multitasking during driving can have detrimental effects on driving a few seconds after the tasks had stopped, which they refer to as “persistence of distraction”.

The assumption that a neural activation triggered by moderate levels of cognitive load could spill over a timely contiguous task with significant effects on performance is speculative as no neurological measurements were taken. Furthermore, most studies with refractory periods or task-switching paradigms reported spillover effects lasting only a few seconds (Pashler, 1994; Rogers & Monsell, 1995) not minutes as in this study. On the other hand, a recent study about nudges suggested a temporal spillover effect aligned with our results, since both studies reported contiguous effects with a longer time-frame (Van Rookhuijzen et al., 2023).

Given the study limitations highlighted above, the reported findings can be taken as preliminary evidence and the starting point for a series of further studies, including tasks with distinct levels of cognitive load and similar durations in scenarios of daily commuting, which could involve immersive scenarios and driving simulators mimicking factual cars. This would allow not only testing if/how the cognitive load triggered by one task can spillover into a timely contiguous task and affect performance, but also examining the proposition that while a moderate cognitive load would be beneficial for performance, higher levels of cognitive load would impair it, as observed in cognitive assessments made after spaceflights (Tays et al., 2021).

Conclusion

To our knowledge, the effects of a demanding cognitive task on driving were only examined when the cognitive task and the driving task were concurrent (cf. Alm & Nilsson, 1995; Jamson & Merat, 2005; Lamble et al., 1999). The effect of prior cognitive load on subsequent driving or the potential effect of the cognitive load demanded by driving on subsequent cognitive tasks had not been investigated before. Despite its limitations, the findings in this exploratory study offer an insight into the effects of cognitive load on timely contiguous cognitive tasks linked to car driving. They suggested that: (i) the completion of tasks with a higher cognitive load before driving can lead to a modest increase in the distance kept from a leading car, and (ii) the completion of a driving task with a higher cognitive load than the control condition before another cognitive task can improve accuracy (but slow reaction time).

The pattern of responses in the two studies points to a rich and complex interaction between cognitive load and driving (with its own cognitive load), which may have practical implications for work performance and driving safety, as pointed out by Vilchez and colleagues (2023). The processes underlying such spillover effects are unknow but could be linked to the activation of long-lasting attentional processes involving alertness. Thus, this study offers a steppingstone towards the development of an experimental framework for future studies investigating the effect of cognitive load in scenarios of daily commuting by car.