1. Introduction
Driving a car is a demanding task [
1] that is known to be problematic for a subset of older drivers due to declines in different cognitive domains [
2]. Therefore, different countries have implemented specific policies to identify unsafe older drivers, and if necessary, to revoke a driving license [
3]. Nevertheless, it has been shown that several of these policies have mostly failed [
4,
5]. However, driving practice interventions, especially for older drivers, may be an alternative approach to counteract reduced driving performance.
Focusing on driving in different age groups, many studies have demonstrated that the type of car accident differs between young and old drivers [
6], and that there is an increased risk for older drivers to be involved in collisions, in more complex traffic situations [
7,
8]. Similarly, in complex traffic situations, older drivers tend to make more driving errors than younger drivers [
9]. Anstey and Wood [
2] reported on complex traffic situations, in which cognitive domains, such as selective attention, switching, inhibition, and discrimination predicts braking/accelerating and lane change errors in older drivers.
These differences between younger and older drivers are also discussed in the context of changes in cognitive decline, and driving safety [
10,
11], unsafe driving behavior in older drivers [
12], and the use of compensatory behavior [
13,
14]. While educational approaches have been shown to raise awareness for problematic driving aspects in the older driver community [
15], they did not increase driving safety in this age group [
16]. Based on this knowledge, driving practice interventions may be an alternative approach to counteract reduced driving performance. In recent years, different training approaches for older drivers have been developed. For example, cognitive training studies have revealed promising training effects on driving safety [
17], a delay in driving cessation [
18], or the enhancement of protective driving behavior [
9,
19]. Roenker and colleagues [
20] revealed that a speed of processing training compared to a passive video-watching in a driving simulator cabin increased driving performance. In addition, practicing visual scanning at intersections (second look) in a driving simulator leads to positive long-term effects, measured two years afterwards (number of second look) [
21]. The reason for the advantages of interactive driving simulator training is likely due to the complexity of engaging multiple cognitive abilities within a comparable environment to real life. This is in accordance with previous research, which has demonstrated cognitive training benefits for older subjects, when practicing complex and strategic computer games (e.g., improved executive functions) [
22].
In this paper, we will use the mental workload concept as a theoretical concept explaining older drivers’ ability to drive a car safely and efficiently. Mental workload is a frequently used concept in the context of traffic research, and particularly in research disciplines examining the efficiency of drivers’ performance. O’Donnel and Eggemeier [
23] define mental workload as “the portion of an individual’s limited mental capacity that is actually required by task demands”. As has been mentioned by several authors, mental workload “emerges from the interaction between the requirements of a driving task, the circumstances under which it is performed, and the skills, behaviors, and perceptions of the pilot” or driver [
24,
25]. In this context, “situational awareness” is also an important concept. Borghini et al. [
24] define situational awareness as the capability of drivers for the “perception of different environmental elements, with respect to time and space, together with a comprehension of their meaning, and the projection of their status after some variable has changed with time.” Mental workload and situational awareness are tightly connected as an increase in mental workload (induced by increased task demands), which most probably leads to a decrease in situational awareness, which in turn will result in decreased task performance. From the above-mentioned definitions, it is clear that very skilled drivers can manage more demanding traffic situations effectively, and with less errors. Thus, improving driving skills will reduce mental workload, and in turn will improve driving performance. The still unanswered question is, which kind of trainings might improve driving and cognitive skills in older and younger drivers.
In the context of aging, mental workload might increase faster and be sustained for a longer time, in more or less normal and average traffic situations, in a subset of older subjects. The reasons for this could be a general age-dependent cognitive decline, less driving skills due to less driving practice, or an unnecessary anticipation of driving problems caused by a negative self-image of the older driver. The latter possible reason is most likely evident in countries where older subjects are explicitly targeted as traffic risks. Therefore, increasing the mental workload capacities of older subjects might have beneficial effects on the driving performance in these drivers. Possible interventions that could be employed to increase mental workload, in the context of driving, could be intensive driving trainings (increase of driving skills), intensive cognitive trainings (increase of cognitive capacities), or optimization of the self-concept to a more self-confident, and realistic view of their own driving performance.
A study that focused on different training methods for older people (single vs. multitasking training) has been recently published by Anguera et al. [
26]. The results of these training approaches (tracking or discrimination task vs. combination of the two tasks), demonstrated stronger changes in behavioral and neural measures in participants conducting multitasking training, compared to single task training. Better task performance (reaction time) and higher frontal theta power were observed in the multitasking training group [
26]. Thus far, it seems that multitasking training approaches show greater benefits or stronger transfer effects to different higher order cognitive domains [
27,
28]. In order to drive correctly, more than one perceptual and cognitive process is needed [
29]. The effective simultaneous usage of several cognitive processes during driving (e.g., multitasking) declines with chronological age and affects the level of mental workload [
30]. Mastering traffic situations that are demanding seems to be a form of multitasking, for which the mental workload increases [
31,
32,
33]. For example, whenever a dual-task exercise is necessary, the mental workload increases, whereas the overall driving performance decreases [
34].
There are some validation studies for driving simulator applications. A high association (
r = 0.716) or model fit (
R2 = 0.657), in a sample of 129 older drivers, between virtual and real on-road driving behavior was found [
35]. In addition, another study demonstrated that driving simulator behavior correlated with on-road driving performance (
r = 0.599) and cognitive performance (
r = 0.474), or a good model fit (
R2 = 0.50) in older participants [
36]. Furthermore, Engstrom and colleagues [
32] have revealed a relationship between mental workload in real and simulator driving. As shown, it is possible to use driving simulators as interactive training instruments for traffic relevant aspects in cognition and/or behavior [
9,
19]. One of the advantages of a driving simulator is the possibility to train multiple cognitive processes, which are needed during active driving in real time, in a controlled laboratory setting.
Cantin and colleagues [
37] compared younger and older participants in their handling of a multitude of different complex traffic situations in a simulator. For both groups, the mental workload increased with the complexity of the driving context. Nevertheless, younger drivers responded more than twice as often to an additional task, drove faster than older drivers, and in general showed a lower mental workload in a complex driving context. Therefore, the authors concluded that in situations requiring a higher mental workload, older drivers use compensation strategies (e.g., slower driving) that lower the mental workload. Furthermore, Bélanger et al. [
29] revealed that older drivers who crashed in simulated overtaking maneuvers reported more “feelings of mental workload” (subjective distress) compared to non-crashers. Additionally, their cognitive test performance was lower in different cognitive domains (processing speed and attention). Correct inhibition and fast discrimination seem to be key processes for better short-time decision making, which is essential for driving. As in the study by Anstey and Wood [
2], reaction time as a dependent variable is not sensitive enough to evaluate driving performance.
For correct and safe traffic navigation, it is of utmost importance that particular mental processes are still operative to a particular degree. To drive a car efficiently, several psychological functions need to be orchestrated. Most prominent of which are executive functions, including inhibition, attention, planning, and working memory (WM). However, it has frequently been argued that, due to old age, inhibitory processes can be impaired, which has led to the inhibitory deficit theory of aging [
38,
39,
40]. In this theory, inhibition is considered to be a mechanism that suppresses ongoing or competing psychological functions. People with poor inhibitory abilities have difficulties with attention control, as well as with memory function [
41], and additionally, with driving a car [
42,
43]. As inhibition capabilities strongly depend on the amount of workload [
44], older people may demonstrate less inhibition when the workload increases. Strategies or trainings adapted to improve inhibition performance could be efficient in lowering the workload and consequently, enhancing driving performance in old age.
A promising neurophysiological measure for mental workload, which currently has only been used in the context of studies examining younger subjects, is the EEG based
brain workload score [
45,
46,
47]. This brain workload score is defined as the ratio between frontal theta and parietal alpha power (theta Fz/alpha Pz), as measured from standard EEG registrations [
46,
48,
49]. Increased task demands, and therefore, enhanced mental workload, is also associated with an increased frontal theta activity and a simultaneous decrease in parietal alpha activity [
46,
48,
49,
50]. Several studies have used this brain workload score during driving situations. Thus far, they have demonstrated that this score changes during cognitive tasks that require more mental workload, and because of increased traffic demands [
51,
52]. Extensive demands on executive functions concern frontoparietal EEG coherence in the alpha and theta bands [
50], and are associated with an increased brain workload score [
49]. In a recent study, brain activity was recorded during driving in a simulator, which involved tasks of varying difficulty (with or without alertness and vigilance tasks). The results showed that the brain workload increased in relation to the level of task difficulty [
51]. Lei and Roetting [
52] observed in a younger sample, different mental workloads (WM: n0, n1 and n2 back) and driving conditions (passive/active, perform lane change during 75 or 100 km/h). During task execution, brain activity was recorded, which revealed that lane change deviation, WM error rate, and response time delay increased with driving task load and WM load. Moreover, brain workload increased with WM load.
In summary, evidence in the literature on mental workload, in the context of driving behavior, suggests that multitasking influences brain activity in older adults [
26], and that the probability of driving errors increases simultaneously with increases in mental workload [
37]. Therefore, brain workload might also relate to driving performance in older drivers. In our first paper from the Drive-Wise project at the University of Zurich, we found a greater benefit in on-road driving for older drivers participating in driving simulator training, compared with an attention-training group. The attention and the driving simulator training groups were both associated with general cognitive improvements, but only the participants from the driving simulator training group improved their on-road driving performance after training [
53].
In this paper, we will report a second finding, obtained from this project. Here we report changed brain workload measures as a consequence of the different training regimes. All participants performed three inhibition tasks, before and after training while EEG brain workload measures were obtained. These psychological tests and the EEG measures took place in the same weeks as the on-road test drives and application of the cognitive test battery. Thus, we obtained performance measures of the inhibition tasks and brain workload measures, before and after the training interventions.
We hypothesize that the driving simulator training will result in more efficient processing of several psychological functions, including alertness, selective attention, as well as inhibition of irrelevant processes. All these processes are highly relevant for effectively and safely driving a car. Since the subjects who practiced driving a car will improve their attention and inhibition skills, brain workload will also substantially diminish when they perform attention and inhibition tasks outside the driving simulator. Thus, we anticipate a transfer from the driving simulator training to the performance in typical executive function tasks. Such kinds of transfer will not be present, or to a much smaller degree, for the cognitive training or the control subjects.
3. Results
The data from all 77 participants obtained during the inhibition tasks were included in the analysis. Due to technical problems during the EEG recording for one participant, this participant was excluded from the final analysis. Thus, the statistical analysis reported here relies on 76 participants.
Training gains on behavioral data for all three paradigms (Stroop, Negative Priming, Flanker) revealed no significant differences between the training groups and the control group, or between the simulator training group and the attention-training group. Descriptive statistics from pre and post inhibition performances are displayed in
Table 1. Compared to the control group, there was no significant linear improvement as a consequence of training, with respect to the reaction time difference in the Stroop task (
F(1,74) = 0.46,
p = 0.50), in the Negative Priming task (
F(1,74) = 0.03,
p = 0.86), and the Flanker task (
F(1,74) = 0.39,
p = 0.53). Planned contrasts between simulator training and attention training revealed no differences in reaction time difference (incongruent minus congruent trials) in the Stroop task (
F(1,74) = 0.30,
p = 0.59), in the Negative Priming (negative minus normal trial) task (
F(1,74) = 0.44,
p = 0.51), and the Flanker task (incompatible minus compatible trial) (
F(1,74) = 0.02,
p = 0.90). Due to these non-significant findings for the behavioral data, and the obvious similarity of the measures obtained during the pre- and post-tests, no effect sizes were calculated (all
d values are approximately 0).
Descriptive statistics of the brain workload changes from pre- and post-training conditions are displayed in
Table 1, including Cohen’s
d for the pre–post differences for the three groups. Interactions between the two specific contrasts, and the linear trend of training gains due to brain workload changes in each inhibition task are displayed in
Table 2 and in
Figure 3. Compared to the control group, there was no significant linear improvement in brain workload (
F(1,73) = 0.88,
p = 0.18) for the Stroop task as a result of the training, but there was a significant linear improvement for the simulator training group compared to the attention training group (
F(1,73) = 3.49,
p < 0.05). Compared to the control group, there was no significant linear improvement in brain workload (
F(1,73) = 1.40,
p = 0.12) for the Priming task as a result of the training, but a trend for linear improvement in the simulator training group, compared to the attention training group (
F(1,73) = 1.68,
p = 0.10). Compared to the control group, there was a trend for linear improvement in brain workload (
F(1,73) = 2.28,
p = 0.068) for the Flanker task as a result of the training, and a significant linear improvement in the simulator training group compared to the attention training group (
F(1,73) = 5.86,
p < 0.01).
4. Discussion
The main goal of this research was to investigate whether the performance in inhibition tasks and brain workload (indexed by theta Fz/alpha Pz) change because of different training regimes. In this paper, we examined the influence of a driving simulator and attention training, on inhibition tasks, performance, and brain workload. Supplementing our already published results on the benefits of driving simulator training [
53], this paper demonstrates that realistic driving simulator training also reduces brain workload during performance of the inhibition tasks outside the driving simulator situation. No changes in the behavioral performance (inhibition task) was found, which emphasizes that fewer brain resources are needed for the same reaction time and error rates. The lack of change in the inhibition task performance can be explained as a consequence of the inhibition task paradigms, since no change in mental workload was included. For this reason, task difficulty did not change. In line with the neural changes in the driving simulator group, the task execution was easier after training than before. Since driving simulator training can be considered as multi-domain training, complex multi-domain trainings tend to exert greater efficacy than single-domain training approaches, in regards to various outcome measures [
26,
66]. In this context, it has also been shown that driving simulator training, as a model for multi-domain training, induces improvements in on-road driving [
9,
33], while single-domain trainings are less effective [
22,
27,
28]. It is important to note that these improvements are associated with a shift in brain workload measures.
Recent publications have demonstrated the importance of complex training approaches, and their relationship to changes in mental and/or brain workload [
26,
66]. Our main finding is that older participants who took part in the simulator training, showed a decrease in brain workload during the performance of tasks requiring inhibitory functions. However, it is important to note that the performance in these tasks did not change, so the simulator training reveals an advantage over the other training regimes, with respect to the fact that less neuronal resources are needed after training, to conduct inhibitory tasks.
Further research could focus on the reasons for this selective influence of the simulator training. Our results support the idea that the driving simulator training requires more brain workload than a consecutive attention training, which has been proven by other authors. Cantin and colleagues [
37] have demonstrated that driving complexity is associated with mental workload in older adults. Also, Lei and Roetting [
52] revealed in a younger sample that brain workload increases in line with mental workload. Handling a driving simulator during different demanding traffic situations entails the efficient usage of several psychological functions (including inhibition of inadequate responses), either sequentially or simultaneously [
29,
30]. While practicing this demanding task, the subjects may learn to allocate the neurophysiological resources more efficiently [
67]. Thus, this finding is in line with previous research, demonstrating that demanding multi-domain trainings tend to have greater efficacy or impact than single-domain trainings [
22,
27,
28]. Moreover, this neuronal shift could be the reason why simulator trainings induce positive behavioral changes in on-road driving [
9,
21,
33], in healthy older subjects.
Although the brain workload measures obtained during the performance of the inhibition tasks decrease as a consequence of the driving simulator training, the performance in these tasks remain unchanged, even for the attention training. The reasons for the lack of training effects on behavioral measures of executive functions in older subjects need further investigation [
68,
69]. It could be that the underlying neurophysiological processes change as a consequence of task-relevant, age-dependent compensational strategies. This idea is supported by a recent finding of Wild-Wall, Falkenstein and Hohnsbein [
43]. Although these authors did not conduct a training study, they reported a type of dissociation between behavioral and neurophysiological indices of executive functions. They described a general slowing in performing executive tasks for the older subjects. This was not accompanied by reductions in interference effects, similar to those found in younger subjects. As the authors identified enhanced frontal N1 ERP amplitudes in older subjects, they suggested that older participants pay more attention to the task without influencing the task performance. Such compensatory neurophysiological activations are frequently found in studies comparing young and old subjects [
70,
71]. They are the pivotal argument for the posterior-anterior shift theory in aging [
71].
Furthermore, it is important to note that the three inhibition tasks are not identical in relation to their theoretical framework. In the Stroop task, participants have to suppress reading. Therefore, participants must consciously suppress an automated brain function [
72]. The Negative Priming task measures another automatic inhibitory process in addition to awareness [
59]. In the current study, none of the participants realized the existence of a distractor, and as a result the Negative Priming effect was observed. The Flanker task requires non-automated brain functions, associated with inhibitory functions, of which the subjects are aware. The task difficulty is based on the fact that the subjects are required to actively suppress salient information with high task relevance [
73]. Nevertheless, there is one common parameter in all three tasks, which is processing speed. Processing speed is frequently described as a typical function, for which older participants demonstrate weaker performance [
74]. This “slowing” in older subjects is often evident in situations when different cognitive functions compete for control resources, which typically occurs in inhibition tasks [
75]. Several studies have reported that older subjects use various compensatory brain resources to cope with these demanding situations [
76,
77,
78,
79]. Another study investigating the inhibitory deficit theory, found a difference in effect between younger and older adults, but no age-related interference effect [
80]. The authors concluded that younger and older adults did not differ in their relative incapability to prevent the processing of irrelevant information. Transforming these findings to our results, we interpret the neuronal changes without any accompanying behavioral change, as a shift in cognitive strategies. Both training approaches included speed-sensitive components (feedback performance of reaction time), but at the behavioral level, no reduction in reaction time difference (RT
Inhibition minus RT
Neutral) was observed. The change in brain workload in the simulator group may represent a shift from compensational, to more efficiently operating inhibitory processes, which needs less mental and neurophysiological resources. Anguera et al. [
26] found comparable results and interpreted their findings as:
“training-induced neuroplasticity as the mechanistic basis of these training effects”. Fronto-striatal circuits [
81] play a crucial role, not only for motor functions, but also for cognitive and emotional functions as well. Tisch et al. [
81] concluded the following on page 770.
“To summarize, of the fronto-striatal circuits described to date, only two, the motor and oculomotor circuits, have primarily motor functions. The remaining circuits are nonmotor and play a role in specific aspects of cognition or in regulation of drive, motivation, mood, and elements of social behavior.”
Based on the current results, some of these circuits might be used during the simulator training more extensively than in the single domain cognition training, due to multitasking, task novelty and complexity. In accordance with Gevins et al. [
47], participants from the simulator training group completed inhibition tasks with equal performance, but needed less attention (reduced frontal theta), and were in a state of relaxation (increased parietal alpha) during task completion. Participants from the simulator group might perform with more flexibility and use less compensatory strategies after training [
76,
77]. In line with the current results, it seems that there are inhibitory fronto-striatal circuits with on-road driving relevance. Further research in this field should focus on these circuits and its influence on ageing and driving performance.
Our hypothesis was that the simulator training would induce psychological and neurophysiological processes, reducing the task-related brain workload level more than single attention trainings. Our hypothesis is grounded by the idea that complex driving simulator training, activates and trains several psychological functions, while consecutive single attention training generally makes use of only one psychological function. Therefore, when using a higher brain workload level for a longer time, the involved neural networks may have the opportunity to adapt to these demands, and develop a more efficient wiring, which at the end causes reduced brain workload levels during demanding tasks in general. In fact, we identified reduced brain workload levels during inhibitory tasks, only for those participants practicing the driving simulator. Further research needs to investigate how these gains affect driving safety, and whether complex training approaches can be implemented as tools for high-risk drivers, or for neurological rehabilitation programs.
5. Limitations
Since we used a priori defined contrasts, we could not test all possible training induced differences statistically. Therefore, we were only in the position to describe some potentially interesting training related differences descriptively.
A major limitation of this type of study is that it is unclear how long the observed beneficial effects of the driving simulator training will persist. There are some papers reporting reduced numbers of collisions [
17], improved driving behavior, or cognitive performance for quite a long period after training [
17,
21,
82]. However, further research is needed to examine the long-term consequences of driving simulator training.
Here we have used a relatively simple brain workload measure (theta Fz/alpha Pz), however, several other brain workload measures have been proposed, most likely reflecting more precisely the neurophysiological resources used, to control particular psychological functions [
24,
67,
83,
84,
85].
It should also be noted that driving simulator trainings are also associated with some disadvantages. A fraction of the subjects suffers from simulator sickness or dizziness. In our study, we tested this explicitly, and excluded affected subjects, who suffered seriously from these disadvantageous states. However, we adapted our simulator setting in order to reduce possible simulator sickness feeling, by using smaller monitors and we disassembled the movable platform underneath the simulator seat that simulated car movements.
In addition, we did not control what the subjects did during the training period, when they were outside of our laboratory. Therefore, it might be that those subjects who had been enrolled in the driving simulator training were those who actually drove more in their car, because they had become more confident in their driving skills. It could also be possible that the specific training scenario stimulated other kinds of behavior, which the subjects perform more frequently outside the laboratory, which might be beneficial for the subjects. However, this is a general problem for almost all long-term training studies published so far, because it is rather difficult to control what the subjects actually do when they are outside the laboratory. Future studies using mobile devices might help to solve this problem.
Finally, we would like to acknowledge that we discussed our findings in the context of cognitive inhibitory control mechanisms, which were improved during the driving simulator training. The multiple factors (e.g., emotion, motivation, or for example personality) and its complexity is discussed in the review published by Nigg [
86]. In our study, we controlled emotional and motivational aspects in our sample [
53] and therefore we were not able to focus on the whole complexity of inhibitory control mechanisms. However, to formulate elaborate hypotheses about possible further influences on involved processes, in the context of inhibitory processes, we would need a different experimental design to delineate these different processes. Hopefully, future studies might be conducted accordingly.