Human behavior in social relationships is prescribed by rules, and it is necessary to follow rules to lead a successful social life. Rules determine the relationship between the value of an action and a specific situation; therefore, humans choose their actions based on a value arising from a rule pertaining to the current situation (Bunge, 2004). When a large-scale advanced information system is operated, such as a power plant or air traffic control center, it is very important for the operator to choose the appropriate behavior within the restrictions engendered by the rules for the situation, which change dynamically to maintain the safety of the system. Tactical thinking is required to follow the rule and determine an appropriate behavior under that rule. In addition, an interaction between the situation and the appropriate behavior are important in that tactical thinking. Cognitive control, or executive function (Miyake and Friedman, 2012), is an essential cognitive function that underlies the placing of a value on a situation, and choosing the appropriate response under a specific rule; the lateral prefrontal cortex (LPFC) plays a key role in cognitive control (Bunge, 2004; Badre, 2008; Dixon, 2015). A previous neuroimaging study suggested that the magnitude of the expected incentive is related to increased LPFC activity reflecting the cognitive control process (Dixon and Christoff, 2012).

Enhanced LPFC activity represents the increase in cognitive load needed to perform difficult experimental tasks. Previous neuroimaging studies using a working memory task showed a robust relationship between LPFC activity and task difficulty (Braver et al., 1997; Manoach et al., 1997; Ayaz et al., 2012; Fishburn et al., 2014; Ozawa et al., 2014). Other studies have suggested that the LPFC plays a key role in attentional control (Rossi et al., 2009), and that activity in the LPFC represents attentional status for a task requiring sustained attention (Derosière et al., 2014). In addition, similar relationships between LPFC activity and difficulty with language processing (Miura et al., 2005) and arithmetic processing (Verner et al., 2013) have been reported. Those relationships are associated with the increased demand placed on cognitive resources as task difficulty increases.

Increased task difficulty, and addition of an extra rule, affect the degree of cognitive control required. However, it remains unclear how cortical activity associated with tactical thinking—to choose an appropriate behavior under the additional rule—affects cortical activity in terms of coping with a difficult situation and the original rule. As mental workload increases after adding a specific rule, the difficulty of task execution may also increase. However, mental workload would also increase when trying to adapt to a limitation engendered by that rule, even if the task is easy to accomplish, and, therefore, may be dissociated from the increased mental workload required with increased task difficulty.

We hypothesized that these two factors affect increased cortical activity individually, and that the effects might be different in the left and right LPFC. To examine this hypothesis, we designed an experimental video game in which the player catches balls falling on a computer screen. Furthermore, to explore the interactive effect between task difficulty and the existence of an additional rule, to determine the outcome on cortical activity, we manipulated the difficulty of the gameplay, and added a rule that affected the score. Functional near-infrared spectroscopy (fNIRS) was utilized to measure cortical activity in the prefrontal cortex in this study. fNIRS measures changes in cortical surface activity based on changes in the relative density of oxyhemoglobin (oxy-Hb) and deoxyhemoglobin (Villringer and Chance, 1997; Ferrari and Quaresima, 2012). fNIRS measures activity under conditions of lower restraint of body motion compared with other neuroimaging techniques, such as functional magnetic resonance imaging and electroencephalography; therefore, the fNIRS technique has been widely used to elucidate cognitive functions (Masataka et al., 2015) in a neuroergonomics study (Derosière et al., 2013) and to develop a brain-computer interface (Bauernfeind et al., 2008; Pfurtscheller et al., 2010; Yanagisawa et al., 2010; Strait and Scheutz, 2014; Naseer and Hong, 2015). From these findings, we expected to measure cortical activity even when the experimental task required continuous hand and arm actions by the player. As a consequence, the present study conducted an fNIRS experiment in which task difficulty was manipulated and a rule related to the experimental task was added to dissociate cortical activities, which increased the task difficulty and modified behavior to cope with the additional rule.

Table 1 summarizes the mean scores obtained for each task, and the ratio between the obtained score and the perfect score for each task. All participants received a perfect score under the Easy without an additional rule condition but scores decreased when task difficulty increased or the additional rule was applied. The two-way repeated-measures factorial ANOVA for obtained score of each task revealed significant main effect of task difficulty (F_(1,19) = 106.8884, p < 0.0000, generalized η² = 0.7195) and the existence of additional rule (F_(1,19) = 16.9618, p = 0.0006, generalized η² = 0.0530), and significant interaction effect was also detected (F_(1,19) = 11.3884, p = 0.0032, generalized η² = 0.0245). A post hoc analysis showed that the factor of the existence of the additional rule showed significant difference at the difficult task condition (F_(1,19) = 19.8718, p = 0.0003, generalized η² = 0.0767), but that difference was not observed under the easy task condition (F_(1,19) = 19,363, p = 0.1801, generalized η² = 0.0485). Therefore, the existence of additional rule influenced decrease of score when the task difficulty was set at difficult level.

Figure 3 illustrates the relative changes in oxy-Hb density during each task compared with the resting period just before the task for each measurement channel. Bilateral LPFCs showed a large increase in signal intensity (for right ROI: mean signal = 0.0535, T₍₁₉₎ = 3.0103, p = 0.0107; for left ROI: mean signal = 0.0767, T₍₁₉₎ = 5.0884, p < 0.0000), whereas the signal from the middle part of the forehead corresponding to the front polar region did not change (mean signal = 0.0147, T₍₁₉₎ = 1.2292, p = 0.3510).

Table 2 shows the mean signal changes of the ROIs corresponding to the bilateral LPFCs, and Table 3 summarizes the results of the ANOVA for the mean signals of the ROIs for each condition. And, Figure 4 shows the time series for the relative changes in oxy-Hb density under each task condition in both ROIs. The three-way repeated-measures factorial ANOVA revealed significant main effects of task difficulty (F_(1,19) = 5.8839, p = 0.0254, generalized η² = 0.0573), existence of an additional rule (F_(1,19) = 9.4296, p = 0.0063, generalized η² = 0.0638), and ROI position (F_(1,19) = 6.4378, p = 0.0201, generalized η² = 0.0131). In addition, significant interaction effects were detected between the additional rule and the ROI position (F_(1,19) = 11.7581, p = 0.0005, generalized η² = 0.0069); a post hoc analysis indicated that the factor of the existence of the additional rule showed significant difference at the left ROI (F_(1,19) = 17.9952, p = 0.0004, generalized η² = 0.1142). And, the factor of the location of the ROI showed significant difference when the task involved the additional rule (F_(1,19) = 14.9463, p = 0.0010, generalized η² = 0.0366). Therefore, the signal change on the left LPFC showed a marked increase when the additional rule was exist on the task, but degree of this effect was different in the left and right ROIs, the effect of the additional rule was not significant on the right LPFC (F_(1,19) = 3.4856, p = 0.0774, generalized η² = 0.0284).

The result of correlation analysis revealed that significant correlations were not observed between the performance data and the mean signal changes of the ROIs. The correlation coefficients under the easy conditions (Easy with BR and Easy with BR + AR) could not appropriately compute because of the ceiling effect of the scores. And, the correlation coefficients under the difficult condition for each ROI did not show the statistical significance (for right ROI: r = 0.3195, T₍₁₈₎ = 1.5994, p = 0.6544 for Difficult with BR; r = 0.2999, T₍₁₈₎ = 1.3335, p = 0.7960 for Difficult with BR + AR; for left ROI: r = 0.3520, T₍₁₈₎ = 1.5954, p = 0.5120 for Difficult with BR; r = 0.4333, T₍₁₈₎ = 2.0396, p = 0.2256 for Difficult with BR + AR).

The behavioral analysis showed that scores decreased with higher task difficulty and when the additional rule was included. Score were lowest under the Difficult with BR + AR condition compared with those obtained under the other conditions. Thus, both factors were affected separately by subjective workload. All participants achieved a perfect score under the Easy without additional rule condition, indicating that they understood the game sufficiently well. However, some participants only captured the red ball under the Easy with BR + AR condition, which consequently decreased their score. The same tendency was observed under the Difficult with BR + AR condition. These results indicate that the participants changed their behavioral tactics to avoid the risk of a large point deduction, even if the condition could not be satisfied. Thus, the change in behavioral tactics to cope with the additional rule affected mental workload and task difficulty.

As for a novel finding of this experiment, the left LPFC showed a larger NIRS signal increase compared with that of the right LPFC when the additional rule existed in the task. A significant main effect of ROI location, and an interaction between the factors of task difficulty and existence of the additional rule, were detected. The post hoc analysis showed significant simple effects, indicating that activation of the left LPFC was enhanced when the additional rule was in place. Although the patterns of fNIRS signal changes while the participants performed the experimental task without an additional rule were almost the same between the left and right ROIs, large signal increases were observed specifically in the left ROI when the additional rule existed. Thus, the specific signal increases in the left ROI were induced by the existence of the additional rule, and this effect occurred independent from the increase in task workload (represented by increased signals in bilateral LPFCs). These results suggest that the cognitive mechanism underlying the changes in behavioral tactics had a different effect on mental workload.

The participant could not miss any of the balls to achieve a high score; therefore, they were expected to catch the balls in order, closest to the limit of the fall. Moreover, if a participant decided that they could not catch a ball under the more difficult task conditions, they did not attempt to catch the ball closest to the limit of the fall but tried instead to catch the second-closest. Thus, the behavioral choice for the task without the additional rule may have been made using simple tactics throughout the task block. In contrast, the participant avoided missing the red ball when the additional rule was present because violation of that rule resulted in a large scoring penalty.

Consequently, the participant occasionally decided not to catch some balls, so they could instead catch the red ball before it fell. Thus, the behavioral choice when the additional rule existed was switched to catching of a ball with a specific color to avoid the penalty. The participant had to judge whether the behavioral choice should be switched from the usual tactic when a red ball appeared to make such a behavioral choice during the task. Thus, the cognitive processing required to modify the behavioral choice to cope with the additional rule increased the mental workload, and the specific increase in signal intensity in the left ROI, for the task with the additional rule, reflects the cortical activation underlying that cognitive processing. McGuire and Botvinick (2010) reported a relationship between the subjective cost of a decision and activity in the left LPFC. Braver et al. (2003) suggested that transient cognitive control is associated with reinforced activation of the left LPFC compared with sustained cognitive control during task switching. Berkman and Lieberman (2010) reported that motivated and approach actions activate the left dorsolateral prefrontal cortex compared with avoidance, regardless of stimulus intensity. Bunge (2004) suggested that the left ventral LPFC is associated with rule retrieval. Taken together, these results indicate that the increased signal intensity in the left LPFC is associated with an increase in the cognitive workload required to switch behavioral tactics to deal with the current situation and the additional rule.

The NIRS results also revealed that signal intensity in bilateral LPFCs increased significantly during the game, and that the signal increase became highest when the difficult with BR + AR condition was performed. Since decrease of the score is considered to reflect the workload for each condition, the difference of signal increase would represent the difference of workload to perform each task condition. Previous neuroimaging studies suggested that the LPFC is associated with cognitive control (Bunge, 2004; Badre, 2008; Dixon, 2015) and has a hierarchical structure (Koechlin et al., 2003; Kouneiher et al., 2009). Activation of the LPFC has also been associated with various cognitive processes, such as response inhibition and task switching (Kane and Engle, 2002; Niendam et al., 2012). Furthermore, the magnitude of the LPFC activation reflects the workload required for the working memory task (Braver et al., 1997; Cohen et al., 1997; Jonides et al., 1997; Manoach et al., 1997; Herff et al., 2014; Ozawa et al., 2014). In addition, a relationship between performance on a working memory task and fNIRS signal intensity from the LPFC has also been suggested (Tanida et al., 2012; Ogawa et al., 2014; Yasumura et al., 2014). Sato et al. (2013) demonstrated the appropriateness of recording the fNIRS signal on the prefrontal cortex and its correlation with the fMRI-blood-oxygen-level dependent signal obtained during a working memory task. Based on these findings, several fNIRS studies have reported that activation of the LPFC can be used as an index of mental workload (Ayaz et al., 2012; Fishburn et al., 2014). Our results support those findings, as the magnitude of bilateral LPFC activation reflected the mental workload to perform the experimental task under each condition, and, consequently, fNIRS would be useful to elucidate the cognitive state of operators who manipulate advanced information systems. Käthner et al. (2014) demonstrated that mental workload level can be determined from event-related potential signals. A method to measure mental workload during an actual situation may be possible by combining these neuroimaging techniques.

From the result of correlation analysis, correlation coefficients under each combination of the score and the signal change on the ROI did not show the statistical significance. It is suggested that the signal increase induced by the task execution on each ROI had large individual variability within each kind of experimental condition. The reason is considered that we manipulated the task difficulty and the existence of additional rule as the experimental factors which influences the mental workload to perform the experimental task, and as a consequence, the signal changes associating with the task performance would be reflected by not only the influence of each factor but also the interaction effect of those two factors. And, it is inferred that a degree of the interaction effect had large individual variability associating with a subjective workload to consider the influence of each factor. Since the subjective workload might be influenced by subjective feeling of time pressure or frequency to change the behavioral choice, the subjective workload would vary with progress of the task. Thus, significant correlations were not observed within each kind of experimental condition. By contrast, ANOVA for the ROI signals and task scores showed significant differences for those factors, and therefore, there was a marked difference of the cognitive load between each experimental condition even though the signal changes of ROIs included large individual variability. It is suggested that the differences of signal changes induced by each experimental condition were larger than the fluctuation of signal change caused by individual variability. Taken together with those results, it could be a supporting evidence that our experimental manipulation could successfully act to distinguish each experimental condition in terms of the cognitive load.

Some limitations of the present study should be mentioned. The present results do not include the peripheral cognitive control processes used to perform the experimental task, such as performance monitoring or risk perception, and those processes are mainly associated with the anterior cingulate cortex (Botvinick et al., 2004; Ridderinkhof et al., 2004) and anterior insula (Paulus et al., 2003; Tom et al., 2007; Rudorf et al., 2012). Because of the characteristics of the experimental task, it was expected that these processes would be involved in task execution. However, fNIRS only measures the surface side of the cerebral cortex, so signal fluctuations deeper in that cortex could not be measured during the experiment.

In conclusion, activity in the left LPFC more closely reflected the increase in cognitive control underlying the behavioral changes needed to cope with the additional rule. This cognitive control affected the change in LPFC activity independent of the increase in task difficulty. Thus, the present results suggest that fNIRS can be used to estimate cognitive processing as a source of mental workload, as well as measure the magnitude of that workload.

NM designed and conducted the experiments, analyzed data and wrote the manuscript. NS designed and conducted the experiments. SK gave technical support and conceptual advice. All authors discussed the results and commented on the manuscript at all stages.

This study was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI (Grant No. 24700121).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.