Activity of Prefrontal Cortex in Teachers and Students during Teaching of an Insight Problem

Participants

Twenty-eight right-handed adults with no medical history of neurological disorders participated in this study (14 pairs; 18 men and 10 women; mean age: 21.5 ± 1.5 years; range: 20–25 years). Participants were paired in same-gender dyads to avoid cross-gender effects (Balliet, Li, Macfarlan, & Van Vugt, 2011). Participants of each pair were randomly assigned to either the teacher or student role during the puzzle-teaching task. The roles of each participant were maintained throughout the study. All participants provided written informed consent, and the protocol used in the study was approved by the local ethics committee (reference no. 2015-1-577).

Teaching Puzzle Task

We utilized a tangram puzzle task as an insight-problem to investigate the PFC activity related to the metacognitive process underlying the teaching and problem-solving processes. A tangram is a puzzle game that consists of seven triangle and square pieces. To solve this puzzle task requires insight; therefore, it can be considered a type of insight-problem. In this task, the teacher gave hints to the student to facilitate solving the tangram task via a tablet screen while the participants were seated diagonally in front of one another to monitor each other (Figure 1). All participants were inexperienced with a tangram game. To familiarize them with this task operation prior to measurement, the participants performed a training task once prior to measurement. We used tangram figures, such as convex polygons, animals, and houses, all of which required wooden tangram parts.

Figure 1 shows the time flow of a puzzle teaching task. A task session consisted of the first 55-s tangram-making phase (prehint phase), a 10-s hint phase (hint phase), and the last 55-s tangram-making phase (posthint phase). After a beep tone, a tangram figure appeared on the tablet screen of both the teacher and the student. Student participants combined wooden tangram parts according to the figure shown during the first 55-s phase. The teacher planned hints while watching the student's behavior as well as the answer of the tangram puzzle, which was shown on only the teacher's tablet screen (Surface Pro 4, Microsoft Corp., Redmond, WA). Next, the word “Hint” appeared on both screens, and the teacher participant directly described two tangram parts in the tangram figure on the tablet screen, using a tablet pen, within 10 s, to help the student perform the tangram task. These hint figures were simultaneously drawn on the student's tablet screen via a local cable and the student then restarted the tangram task for 55 s after the word “Hint” disappeared. Under control conditions, the teaching participants traced the figure's outline during the 10-s hint phase. If the student participants solved the tangram task within the limited time period, they released their hand from the desk and stared at the tablet screen. We used PowerPoint 2016 (Microsoft Corp., Ltd.) so that hints drawn on the teacher's tablet could be seen in real time on the student's tablet (On-Lap 1303, GeChic Corp., Taichung City, Taiwan) via an HDMI cable. The session began with a rest condition of approximately 60 s, followed by a task condition of 120 s, which was followed by another approximately 60-s rest condition.

Participants were instructed to keep their faces turned to the game screen and to minimize head movements under all conditions. Each session included three tasks, and teaching and control sessions were randomly performed twice. The order of the 12 tangram figures of teaching and control conditions was pseudorandom. We evaluated whether the tangram task was completed within the limited time using a video camera.

NIRS Measurement

Two wearable 16-channel NIRS systems (WOT, Hitachi Co. Ltd., Tokyo, Japan) were used to evaluate activation in the PFC while the participant pairs engaged in the teaching puzzle task. A portable processing unit for controlling the optical topography measurements was connected to the probe unit through a flexible cable bundle. The processing unit sent data to a personal computer that controlled the experiment through a wireless local area network. Figure 2 depicts the schematic for the NIRS probes and channels. The NIRS system used in this study consisted of six emitters and six detectors, resulting in 16 channels, each consisting of one source-detector pair. The distance between the source and detector probes in each channel was set to 3.0 cm. The lowest probes were positioned along the Fp1–Fp2 line, according to the international 10–20 system used in electroencephalography. Changes in the concentrations of oxygenated (oxy) and deoxygenated (deoxy) hemoglobin (Hb) were calculated according to the absorbance change of 705-nm and 830-nm light according to the modified Beer–Lambert law (Delpy et al., 1988; Maki et al., 1995). Changes in oxy-Hb values were considered as indicators of changes in regional cerebral blood volume, as oxy-Hb is more sensitive than deoxy-Hb when measuring blood flow changes associated with brain activation (Hoshi, Kobayashi, & Tamura, 2001). We manually marked the start position of each task in the NIRS data from the NIRS software at the time of the start-beep sound produced by the PowerPoint application.

NIRS Data Analysis

The sampling frequency for the NIRS data was 5 Hz. A moving-average filter, with a time-window of 5 s and a band pass filter of 0.2-Hz low pass and 0.01-Hz high pass were used to remove slow drifts and high-frequency fluctuations, respectively. We defined 30 s prior to the onset of the puzzle teaching task, 120 s of the task, and 30 s after task completion as an analysis block. Data from the analysis blocks for each participant were averaged for statistical analysis.

One drawback of the NIRS method is the variability of the path length, which is dependent on the structure of the scalp and superficial tissues over the brain (Haeussinger et al., 2011; Heinzel et al., 2013). To avoid this issue, the oxy-Hb data from each channel of each participant were normalized using a linear transformation, such that the mean ± standard deviation of the oxy-Hb levels in the 10–20 s prior to the teaching puzzle task was 0 ± 1 (AU). This normalization was also useful for circumventing the influence of differential path-length factors between participants and between cortical regions (Harada, Miyai, Suzuki, & Kubota, 2009). To offset the low spatial resolution of NIRS as well as interindividual anatomical variability, the four channels over the right PFC, over the middle PFC, and over the left PFC were averaged, respectively (Figure 2).

We used the Platform for Optical Topography Analysis Tools (Hitachi Corp.) and MATLAB (MathWorks, Natick, MA) software to analyze the NIRS data. Baseline values were defined as the mean of the data recorded 10–25 s before the task onset. Pre-hint1, Pre-hint2, Pre-hint3, Post-hint1, and Post-hint2 values were defined for statistical analysis as the mean of the data recorded at 5–20, 20–35, 35–50, 70–85, and 85–100 s after the task onset, respectively. Pre-hint1, Pre-hint2, and Pre-hint3 were assumed to be the period of solving the task for students and the period of planning how to teach while seeing students' performance for teachers. Post-hint1 and Post-hint2 were assumed to be the period of solving the task using the hint for students and the period of monitoring students' performance using the hint for teachers. Drawing-hint data were defined for statistical analysis as the mean of the data recorded 55–70 s after task onset, and we evaluated the difference in PFC activity related to the motor process when drawing the figure between the control and teaching conditions. If the tangram task was completed before hints were given, we excluded the trial from statistical analysis to focus our evaluation on the role of the PFC in metacognitive functions related to the teaching process. To analyze the neural activity that depended on students' performance after a hint, we compared the mean of NIRS data for the trials when students solved the task following a hint with that when they did not solve the task in the teaching condition.

Statistical Analysis

The mean number of trials of completed tangram tasks was compared between control and teaching conditions using Student's t test. A three-way repeated-measures analysis of variance (ANOVA) for NIRS data was used to determine the effects of Site (right PFC, middle PFC, and left PFC), Period (Baseline, Pre-hint1, Pre-hint2, Pre-hint3, Post-hint1, and Post-hint2), and Condition (Control and Teaching). A two-way repeated-measures ANOVA for NIRS data of the teacher at the Drawing-hint was used to determine the effects of Site and Condition. A three-way ANOVA for NIRS data according to the students' performance after a hint was used to determine the effects of Site, Period, and Task (Completed and Noncompleted). A post-hoc analysis was performed using Bonferroni's correction.

Tangram Task

The mean number of trials of completed tangram tasks in the teaching condition (2.8 ± 1.8 trials) was significantly greater than that in the control condition (1.0 ± 1.2 trials, p = .004). However, there was no significant difference in the mean number of completed tangram trials before the hint between the teaching (0.5 ± 0.5 times) and control conditions (0.7 ± 0.8 times). These results indicate that the hint provided by the teacher was effective in helping students solve the tangram task. As described in the Methods section, we excluded the NIRS data of completed trials before a hint from the statistical analysis.

NIRS Data in Control and Teaching Conditions

Figure 3 depicts the NIRS data from both teachers and students. For teachers, the three-way repeated-measures ANOVA for NIRS values showed a significant effect of Site (F_2,26 = 6.906, p = .004), Period (F_5,65 = 23.171, p < .001), interaction among Site and Period (F_10,130 = 3.052, p = .002), interaction among Period and Condition (F_5,65 = 3.656, p = .006), and interaction among Site, Period, and Condition (F_10,130 = 2.533, p = .008), but no significant effect of Condition or interaction among Site and Condition. In the teaching condition, post-hoc testing revealed that the NIRS values of Post-hint1 were larger than the Baseline value in the left PFC (p = .007). The NIRS values of Pre-hint2 were less than that the Baseline value in the middle PFC (p = .028). The NIRS values of Pre-hint3 were less than that the Baseline value in the right (p = .001), middle PFC (p = .002), and left PFC (p < .001). A two-way repeated-measures ANOVA for Drawing-hint NIRS values showed no significant effect of Site, or Condition, or interaction among Site and Condition. These results indicate that there was no significant difference in the PFC activity of teachers while drawing figures between the control and teaching conditions.

For the students, the three-way repeated-measures ANOVA of the NIRS values showed a significant effect of Period (F_5,65 = 3.504, p = .007), but no significant effect of Site, Condition, or any interaction. Post-hoc testing showed no significant differences.

NIRS Data of the Teaching Condition in Completed and Noncompleted Trials

One participant could solve all tasks and two participants could not solve any of the tasks in the teaching condition. Therefore, we excluded these three teacher-student pairs and analyzed 11 pairs to investigate the change of neural activity depending on student's performance after the hint. Figure 4 depicts the NIRS data of the teaching condition in completed and noncompleted trials. For teachers, the three-way repeated-measures ANOVA for NIRS values showed a significant effect of Period (F_5,50 = 19.138, p < .001) and interactions among Site and Period (F_10,100 = 5.125, p < .001) and among Period and Task (F_5,50 = 3.494, p = .009), but no significant effects of Site, Task, and no interactions among Site and Task or among Site, Period, and Task. Post-hoc testing showed no significant differences in NIRS data between completed and noncompleted trials.

For students, the three-way repeated-measures ANOVA for NIRS values showed a significant effect of Site (F_2,20 = 4.331, p = .027) and an interaction among Period and Task (F_5,50 = 5.602, p < .001), but no significant effect of Period, Task, and no interactions among Site and Period, among Site and Task, or among Site, Period, and Task. Post-hoc testing revealed that NIRS values of completed trials were larger than those of noncompleted trials in the left PFC at Post-hint1 (p = .031).

PFC Role in the Teaching and Problem-Solving Processes

Teachers need not only to control but also to monitor their own teaching strategy according to students' behavior and learning state (Rodriguez, 2013; Watanabe, 2013). Convergent evidence from a number of studies has suggested a critical role for the rostrolateral PFC in these metacognitive processes (Fleming & Dolan, 2012; McCaig et al., 2011). Therefore, we hypothesized that the PFC of teacher participants would activate during the metacognitive processes in both the prehint and posthint phases. However, the PFC activity of teachers increased only after teaching, at the posthint phase, but not while planning hints at the prehint phase. These results indicate that the PFC of teachers might play a metacognitive role in monitoring their own teaching results rather than in the planning teaching strategy based on the students' state. These interpretations are consistent with those of a previous study that found that the PFC functions of teachers are related to monitoring the information regarding their own teaching results and the students' understanding (Takeuchi et al., 2017). Alternatively, activation of the PFC in teachers after giving hints might result from the motor process involved in drawing the figures given as hints. If the motor process itself contributes to activation of the PFC after hints, it is assumed that the PFC activity while describing hints might be different from when tracing tangram figures during the control condition. However, there was no significant difference in PFC activity while drawing figures between the teaching and control conditions. Therefore, it is unlikely that the activation of the teachers' PFC after giving hints resulted from the motor process related to drawing hints.

In contrast to our hypothesis, PFC activity in teachers decreased when they planned their own teaching strategy at the prehint phase. fNIRS studies have shown PFC deactivation during cognitive tasks such as game playing (Matsuda & Hiraki, 2006), verbal memory tasks, when subjects listen to music (Ferreri, Aucouturier, Muthalib, Bigand, & Bugaiska, 2013), and working memory task with a pharmacological stimulant (Ramasubbu, Singh, Zhu, & Dunn, 2012). However, to our knowledge, no fNIRS studies have reported PFC deactivation when teacher participants plan their own teaching strategy. One explanation for this PFC deactivation might be an increase in the attentional mechanism relying on visual stimuli. Previous neuroimaging studies have reported that deactivation of the PFC is related to attention to visual stimuli (Geday & Gjedde, 2009; Matsuda & Hiraki, 2006). The PFC activity of teachers might decrease during the prehint phase as they pay more attention to the tangram figure on the tablet's screen to plan hints for the students. However, in contrast to this hypothesis, a considerable amount of literature supports the importance of the PFC in attentional processes, mainly for maintaining the behavioral goal and protecting them from distracting information (for review, see Ptak, 2012). Therefore, the physiological mechanism of deactivation in the PFC during the planning of teaching strategies remains unclear; nevertheless, our results indicate that planning of a teaching strategy may be less demanding of the PFC.

In the students, there was no significant activation of the PFC before and after receiving hints. However, the PFC activity when the insight-problem task was solved after a hint was greater than that when task was not solved. Once an insight occurs, the PFC can bring the higher cognitive functions to bear on the problem, including central executive processes, such as directing and sustaining attention (Aziz-Zadeh, Kaplan, & Iacoboni, 2009; Dietrich, 2004). A previous study reported that solving a spatial puzzle via insight activates the PFC, which may be important for the metacognitive components of insight solutions, including attention to and monitoring of the solution (Rosen & Reiner, 2017). Therefore, the PFC of students might activate for metacognitive processes in monitoring what they are capable of doing in the case where hints help them to solve insight problems. In contrast, such PFC activation may be absent when students did not achieve insight.

In addition to the metacognitive process, the left PFC might activate because of creativity after hints. The role of the PFC in creativity has been well discussed previously (Dietrich, 2004; Gonen-Yaacovi et al., 2013). However, several studies reported that activation of the right PFC, unlike left PFC in our study, contributes to creativity when solving insight-problem tasks by oneself (Aziz-Zadeh et al., 2009; Rosen & Reiner, 2017; Sheth, Sandkuhler, & Bhattacharya, 2009). These inconsistent results may be because of the differences in experimental and analysis procedures, but the pattern of PFC activation in our study might have been influenced by the difference in the cognitive process involved in solving a problem by oneself and solving a problem with the help of hints. In addition to the creativity process of student for insight-problem solving, the cognitive process of student after receiving teaching has to take account of not only one's own behavior, but also the teacher's hints, which is unnecessary when solving a problem by oneself. Therefore, the explicit metacognitive process for solving insight problems after teaching might result in left PFC activation that differs from the right PFC activation that occurs when solving the problem by oneself.

Limitations

Several limitations should be considered when interpreting the results of the present study. First, the lack of adjustment for the difficulty of the teaching puzzle task per subject may have resulted in a high level of variability in problem-solving performance. Moreover, the mutual interaction between teachers and students was limited, because the duration of teaching was as short as 10 s and the teaching via the tablet was unnatural. Second, the number of participants and task trials in this study were small. Moreover, the participants were not real teachers or students. Third, individual anatomical differences may also have influenced variability in PFC activity. The path-length of near infrared light as well as NIRS sensitivity are dependent on the scalp-to-cortex distance (Haeussinger et al., 2011; Heinzel et al., 2013). Although the NIRS data from each channel of each participant were normalized by linear transformation to circumvent these issues, future studies should address the impact of anatomical differences arising from the cortical, frontal sinus, and skull thickness by using additional imaging methods (Haeussinger et al., 2011; Heinzel et al., 2013). Fourth, the necessity to adjust the teaching strategy according to the student's behavior might be low because of the simplicity of planning hints in our task. Therefore, the PFC activity of the teachers did not increase when planning their own teaching strategy. A future study should evaluate whether the PFC activity of the teacher alters according to the difficulty of planning the hints. Finally, extra-cortical physiological responses, such as blood pressure, heart rate, and skin blood flow should be monitored to ascertain and account for the influence of these parameters on NIRS measurements (Erdogan, Yucel, & Akin, 2014; Kirilina et al., 2012).