Fourteen college students (all males, mean age: 23.4, age range: 21–28), recruited from National Chiao-Tung University, Hsinchu, Taiwan, participated in the current study. This study was carried out in accordance with the recommendations of Taipei Veterans General Hospital with written informed consent from all subjects. The Institutional Review Board of the Taipei Veterans General Hospital approved the protocol. All right-handed participants had normal or corrected-to-normal vision and had driving experience of more than 1 year. The consumption of caffeine, tobacco, alcohol and drugs was prohibited before the experiment. They received a monetary reward for their participation upon finishing the entire experiment.

Experiments for this study were conducted in a driving simulator. A real car was mounted on a dynamic 6-degrees-of-freedom motion platform, and 360° scenes were rendered by seven LCD projectors. In this immersive virtual reality scenario, the participants cruised at a fixed speed of 100 km/h on a highway throughout the experiment. Two tasks, lane-keeping and mental calculation, were utilized in this study.

In the lane-keeping task, the car linearly drifted either to the right (curb) or left (opposite lane) side while cruising in the third lane, as shown in Figure 1A (Huang et al., 2009). Each left or right drift can be considered as one deviation. The participants were instructed to direct the car back to the cruising lane (the third lane) quickly by turning the steering wheel after detecting the random deviation (Figure 1B). There was one sensor to monitor the angle of steering wheel and stored the changes of angles.

The other task in this experiment was the mental calculation task, which involved working memory and induced an increase in the workload (Hitch, 1978; Sammer et al., 2007). In this study, the calculation task was to validate an arithmetic equation as being correct or incorrect (Figure 1C). Here, the ratio of correct to incorrect equations was 1:1. Each equation was based on addition without carrying, and the difficulty level was manipulated to be similar across equations. Each equation was continuously displayed on the screen for 1200 ms. There were two buttons built on the steering wheel for responding to the numerical equation. The right (left) button was pressed with the thumbs if the equation was correct (incorrect).

The deviation and arithmetic equations were displayed simultaneously in the drive (D) and math (M) condition as shown in Figure 1D. To induce distinct attentional requirements, a 400-ms SOA was applied in this experiment. Jensen et al. (2001) have shown that humans require approximately 300 ms (P300 activity) to perceive a visual stimulus, and therefore, an SOA of 400 ms between two vision-based tasks is sufficient for participants in this distracted driving experiment. In the M/D condition, the numerical equation was displayed first, and the deviation appeared 400 ms after the onset of the mental calculation (Figure 1E). In the D/M condition, the deviation was displayed 400 ms before the onset of the arithmetic equation (Figure 1F).

A total of five conditions, two single tasks (D and M), one simultaneous dual task (D&M) and two dual tasks with a 400-ms SOA (M/D and D/M), were randomly displayed throughout the experiment. The inter-trial interval was between six to eight seconds. To ensure the participants’ awareness, they were instructed to undergo normal sleep, and the experiment was executed in the early morning (9:00 AM). Before the experiment, all participants were acquainted with the virtual-reality-based driving environment and the operation of the controls (turning the steering wheel and pressing the buttons) in two training sessions. After that, every participant was required to finish four 15-min sessions separated by 10-min breaks to prevent drowsiness during the experiment.

The behavioral data, including the RT to the deviation and the solution time (ST) of the mental calculation, were recorded. There was a sensor in the steering wheel to collect the turning angle for the reaction to the deviation and a button to record the ST for providing an answer to the mental calculation. All information from the steering wheel was also synchronized with the designed scenario to provide real-time and suitable feedback to the participants.

EEG data were acquired by a 32-channel EEG cap (NuAmps, Compumedics NeuroScan, USA) at a sampling rate of 500 Hz with 16-bit precision. Thirty Ag/AgCl electrodes were arranged according to the modified international 10–20 system, and two reference electrodes (A1 and A2) were placed on both mastoid bones. The impedance between the electrodes and cortex was under 10 kΩ with a NaCl-based conductive gel.

Figure 2A shows the distribution of the RTs (red bars) and STs (blue bars) while only performing the lane-deviation and mental calculation tasks, respectively. The RTs of the driving task were in the range of 440–1188 ms. The participants required more time for the mental calculation, and the distribution of the STs was in the range of 834–2992 ms. Figure 2B shows the averaged RTs and STs, and the STs are significantly longer than the RTs (p < 0.05). The RTs in the dual-task conditions (D&M, M/D and D/M) were not significantly different from those in the single-task conditions. In contrast, the STs in all three dual-task conditions were longer than those in the single-task condition (p < 0.05). The accuracy for responding to both tasks in this experiment was 96.7 ± 1.4%. The differences in the error rates between the single- and dual-task conditions were not statistically significant.

Figure 5 shows the theta and alpha oscillations as a function of time in the frontal, parietal and occipital components while performing the D&M (green lines), M/D (yellowish brown lines) and D/M (pink lines) conditions. The black lines represent the sum of the brain activations induced by executing deviation and calculating. The O-Add, Add and U-Add activations induced by performing the dual-task conditions represent the activations higher than, equal to and lower than the summed activation, respectively. The comparison between the recorded activation and the simulated activation are marked (Wilcoxon signed-rank test, FDR-adjusted p < 0.05).

Figure 5A depicts the comparison of the theta activation. In the frontal area, the O-Add activation is only observed in the D&M condition, and Add activation is found in the M/D and D/M conditions. The Add activation of the parietal theta oscillation is observed in the M/D and D/M conditions, but U-Add parietal theta activation is only found in the D&M conditions. In the occipital region, the Add theta activation is first found, followed by the significant U-Add theta activation.

Figure 5B shows the comparison of the alpha activations in the frontal, parietal and occipital components. The Add frontal alpha activation is observed in all three dual-task conditions. In the parietal and occipital components, the significant U-Add alpha activation dominates during the D&M, M/D and D/M conditions. In addition to the findings in Figure 5, Table 2 summarizes the interactive theta and alpha activations in these three dual-task conditions.

In Figure 5A and Table 2, a non-significant O-Add frontal theta activation was only observed in the D&M condition. In Jaeggi’s study, slight O-Add activation during the simultaneous performance of two working memory tasks (1-back and 2-back) was primarily found in the frontal cortex, which is involved in cognitive processes and the management of concurrent processes (Jaeggi et al., 2003; Dux et al., 2006). The O-Add frontal theta activation in the D&M condition suggests that extra activation was required to perform the tasks. Enhanced frontal theta oscillations have been positively correlated with increased task demands and workload (Klimesch, 1999; Lin et al., 2011; Borghini et al., 2014) as well as decision-making, orienting, conditioning, numerical calculation, workload, working memory and the maintenance of goal states (Gevins et al., 1997; Klimesch, 1999; Sauseng et al., 2005; Kahana, 2006; Sammer et al., 2007; Tombini et al., 2009; Huang et al., 2013). Compared with performing the lane-deviation or mental calculation tasks alone, the extra demands due to attention switching and decision making must be expected. However, Add frontal theta activation was observed in the M/D and D/M conditions (Figure 5 and Table 2). Since a 400-ms SOA was applied in the M/D and D/M conditions, the participants might have been able to optimize their performance in the lane-deviation and mental calculation tasks without extra attentional requirements. The different findings in the frontal theta represent the distinct attentional requirements in three designed dual conditions. These non-significant O-Add or Add frontal theta activations suggest that the extract attention are slightly required as managing lane deviation and mental calculation simultaneously. It means the degree of difficulty of the designed tasks are medium.

Furthermore, U-Add parietal theta activation was only found in the D&M condition after the event onset, but Add parietal theta activation was found in the M/D and D/M conditions (Figure 5 and Table 2). Enhanced parietal theta activation has been shown to modulate the strength of the sensory (Raghavachari et al., 2006). In the current study, two visual stimuli were displayed simultaneously to the participants, and the tasks were displayed sequentially in the M/D and D/M conditions with a 400-ms SOA. The U-Add and Add parietal theta activation suggests that the two visual stimuli can be processed with time-sharing resource allocation (Nijboer et al., 2014). Therefore, U-Add parietal theta activation was expected in the D&M condition.

In the occipital region, Add activation and U-Add activation were both observed in the dual-task conditions. In addition to the time-sharing requirement discussed above, Add activation could also be explained by a specific brain oscillation only induced by one of the tasks (Salvucci and Taatgen, 2008). Increased occipital theta activation was only observed in the early stages of mental calculation (Figures 3B,D), and occipital theta activation has been responsible for controlling cognitive demands (Raghavachari et al., 2006). Specifically, identifying and processing stimuli is associated with occipital activation (Cartier et al., 2012). The increased occipital theta oscillation at the beginning of the mental calculation can be interpreted as a high cognitive demand in response to visual stimuli. Since the increased occipital theta activation was elicited by mental calculation alone (Figure 3B, Table 1 and Figure 4), Add activation was expected. In particular, this occipital theta oscillation was clearly observed after the onset of the mental calculation in the M/D and D/M conditions (the red dotted lines in Figures 4B,C). Following the Add occipital theta activation, a significant U-Add theta activation was then observed in the occipital component. The occipital theta activation was only slightly higher than that observed during the performance of the lane-deviation task (Figure 3). During the dual-task conditions, the U-Add theta activation represents lower resource requirements. This U-Add occipital theta activation is similar to the U-Add alpha activation in the posterior region.

In Table 2, non-significant Add alpha activation was observed in the frontal component, and significant U-Add alpha activations were found in the posterior region (Wilcoxon signed-rank test, FDR-adjusted p < 0.05). Like the theta activation in the frontal area, the posterior alpha is also associated with cognitive processes and attentional demands (Klimesch et al., 1998; Foxe and Snyder, 2011; Lai et al., 2012; Lin et al., 2012a). Due to resource competition between tasks, U-Add activation occurs when one task dominates most of the resources used (Anderson et al., 2011). Decreased posterior alpha activation is associated with visual scanning (Fu et al., 2001). In the current study, reacting to the lane deviation suppressed posterior alpha activation (Figure 3A). Participants were required to continuously focus on the vehicle position with respect to the center of the cruising lane during the lane-keeping task. The posterior alpha suppression could also be attributed to an increased workload and short-term memory requirements during mental calculation (Gevins et al., 1997; Meeuwissen et al., 2011). While performing the mental calculation, participants visually scanned the equation and retrieved numerical meanings to obtain the correct answer for the current equation. Compared with the lane-keeping task, mental calculation is a task that requires more cognitive functions and induces a higher workload. In particular, the averaged STs were significantly longer than the averaged RTs. During the simultaneous performance of the lane-deviation and mental calculation tasks in the dual-task conditions, paying more attention to the mental calculation led to the significant O-Add activation in the posterior area.

In the frontal region, an Add alpha activation was primarily observed in the dual-task conditions (Figure 5 and Table 2). As discussed above, the brain oscillation related to only one task leads to Add activation. Increased frontal alpha activation has been correlated with internally directed attention (Travis and Shear, 2010). Previous working memory studies have reported that increased frontal alpha activation involves in accurately retrieving memorized information (Klimesch et al., 2005; Huang et al., 2013). In the current study, the increased alpha in the frontal cortex was induced while performing numerical calculations (Figures 3B,D). This increased frontal alpha oscillation before the button-press response could be attributed to the fact that participants had to memorize the displayed numerical equation and compare it with the answer while making a mental calculation. The Add frontal alpha activation was expected since the increased frontal alpha activation was only involved in one task (mental calculation).