Taking another’s perspective promotes right parieto-frontal activity that reflects open-minded thought

ABSTRACT

Critical thinking (CT) is important for consensus building. Although the practice of CT using debate is widely used to improve open-minded thought, the cognitive processes underlying this improvement remain poorly understood. Here, functional magnetic resonance imaging (fMRI) was employed to assess how neural responses while considering another’s opinion are changed by CT practice and to determine whether the cortical network showing activation changes related to personality traits is relevant to consensus building. A total of 52 healthy participants were divided into three groups for an intervention; the participants read another’s reasoning regarding a controversial issue and judged whether this person’s viewpoint was affirmative during fMRI measurements. The intervention required them to prepare speech texts from a designated viewpoint based on both themselves and others. Compared to the control group, the group who took another’s perspective showed enhanced activation of the right parieto-frontal network, that has been implicated in belief update, cognitive reappraisal of emotion, and self-perspective inhibition. Additionally, activation of the orbitofrontal cortex was negatively correlated with a stubbornness index. The results provide the first neural evidence of the effects of CT practice and support the notion that open-minded thought underlies the benefits of this practice.

KEYWORDS:

• fMRI
• right parieto-frontal network
• open-minded thought
• critical thinking
• debate practice

Introduction

Modern society is rife with conflict between individuals and/or social groups and it is increasingly difficult to reach consensus. To address these issues constructively in a plural and democratic society, critical thinking (CT), which is defined as rational reflective thinking focused on deciding what to believe or do (Ennis, 1996; Paul & Binker, 1990), has been argued to be an important skill for citizens to possess (Organisation for Economic Co-operation and Development (OECD), 2005; Ten Dam & Volman, 2004; Weinstein, 1993). Open-minded thought, defined as a willingness to revise and reconsider one’s view (Hare, 1979), is an important part of CT as it enables an individual to consider the opinions of others without prejudice (Ennis, 1996; Paul & Binker, 1990; Siegel, 2009). In other words, a good critical thinker should be able to utilize open-minded thought to interpret the opinions of others fairly, to reach a rational conclusion.

Although it has been suggested that debate practice is an effective procedure for improving open-minded thought, the neural underpinning behind changes in open-minded thought remains unclear. Previous studies have suggested some beneficial effects of debate practice including tolerance toward different viewpoints, analytic or communication abilities, self-confidence, and mastering the educational content (Miri, David, & Uri, 2007; Oros, 2007; Hall, 2011; Zare & Othman, 2013; Mumtaz & Latif, 2017). However, those results were based on post-hoc questionnaires or interviews that relied on participant introspection and, therefore, they provide no objective information on how participant mind-set is changed through CT practice. Thus, a neuroimaging experiment would be helpful for understanding changes in mind-set during experimentation following debate practice.

Furthermore, it also remains unclear whether the effects of CT practice exerted any influence on the brain network that explains behavioral tendencies relevant to consensus building. Previous studies have reported correlations between cortical activation and personality characteristics that are relevant to open-minded thought, such as a ruminative temperament (Ray et al., 2005) and a tendency to accept an unfair proposal (Tabibnia, Satpute, & Lieberman, 2008). However, there is no direct evidence regarding whether the brain regions associated with a change in mind-set following debate practice are within the same network as the regions associated with personality characteristics associated with the preference for or avoidance of working on consensus building and discussion, such as a tendency to persist in one’s own perspective.

If a change in mind-set reflects a progression in open-minded thought, one would expect that the cognitive processing accompanying the revaluation of target information would enhance such thinking. Given that open-minded thought involves readiness to reconsider others’ views (Hare, 1979) and debate practice can enhance tolerance toward different viewpoints (Miri et al., 2007; Oros, 2007; Hall, 2011; Zare & Othman, 2013; Mumtaz & Latif, 2017), debate practice should cultivate a mind-set that does not deny others’ opinions because it might be necessary to reinterpret target information and to regulate emotional responses to such information. The following cognitive processing can be considered as possible candidates that underlay revaluation processes. The first involves a revaluation of beliefs, which refers to a learning process for updating the causal associations underpinning one’s own knowledge and obtained information (Corlett et al., 2004; Fletcher et al., 2001). The second involves a cognitive reappraisal of emotion, an emotional regulation strategy that focuses on an antecedent event (Gross, 1998). The last involves the inhibition of a self-perspective because the participant’s own beliefs must be bracketed to allow for fair consideration of others’ perspectives. Previous neuroimaging studies have found that these cognitive processing strategies involve activation of the right parieto–frontal network (Fletcher et al., 2001; Lévesque et al., 2003; Corlett et al., 2004, 2006, 2007; Turner et al., 2004; Ochsner & Gross, 2005, 2008; van der Meer, Groenewold, Nolen, Pijnenborg, & Aleman, 2011; Hartwright, Apperly, & Hansen, 2012, 2015; Ochsner, Silvers, & Buhle, 2012; Buhle et al., 2014; Sugiura, Seitz, et al., 2015); therefore, we expected that the progression in open-minded thought would enhance activation of the right parieto–frontal network.

The purpose of this study was two-fold. The first objective was to determine whether the neural responses observed while considering the opinion of other individuals would change after taking another’s perspective. We hypothesized that the neural substrates underlying open-minded thought would be active after considering another’s viewpoint, while they would not be active when individuals take their own perspective in the debate practice. To test this hypothesis, we designed an experiment consisting of debate practice as a CT intervention and functional magnetic resonance imaging (fMRI) measurements before and after the CT intervention. The CT intervention required participants to prepare a constructive speech, which is a major element of debate practice, from a designated viewpoint. Writing a constructive speech from a viewpoint contrary to that of his/her original perspective would inevitably lead to open-minded thought because the argument must be built by accessing the perspective of the other. By contrast, the perspective of another is not necessarily accessed when participants try to make a constructive speech based on their own original viewpoint. Thus, cognitive workload and control of emotional response during writing might be elements of practice, but differences in designated viewpoints would particularly affect the promotion of open-minded thought. An fMRI task required participants to interpret the opinions of others with regard to several contentious issues. Although the task did not explicitly require the participants to consider the perspective of another person, the neural responses reflecting open-minded thought would be enhanced if the participants performed the CT intervention from the viewpoint of another person. In addition, enhanced activation within the right parieto-frontal network would be expected if the observed neural responses truly reflected the promotion of open-minded thought.

The second purpose was to clarify whether cortical activation reflecting open-minded thought was associated with a stubborn personality. Because the tendency to persist in one’s own perspective would inhibit rational thinking and reaching consensus building, this tendency would be the opposite of open-minded thought. Thus, the present study measured stubborn personality with a “power to live” questionnaire (Sugiura, Sato, et al., 2015). In this questionnaire, the “stubbornness index” measures how this personal trait promotes persistence of one’s own belief (example of items: I am stubborn and always get my own way). By contrast, there were no test measures commonly used to depict directly the personality of sticking to one’s own perspective, though some measures listed stubbornness as one of the item explanations. Thus, the stubbornness index could directly measure a characteristic that opposes open-minded thought.

Materials and methods

Participants

Fifty-two healthy Japanese volunteers (39 males and 13 females, mean age = 21.3, SD = 1.6, range: 19–24 years) participated in this study. We conducted a power calculation based on our experimental design and desired power (0.8 at alpha p = 0.05) prior to the experiment. The power calculation showed that 42 participants were needed to detect a medium effect size (Cohen’s f = 0.25 at alpha p = 0.05 (two-tailed)). Participants were recruited to satisfy the power calculation while also considering data exclusion. None of the participants had a history of neurological or psychiatric disorders. Participants were judged to be right-handed according to the Edinburgh Handedness Inventory (mean laterality quotient: 0.98; Oldfield, 1971). All participants provided written informed consent in response to an experimental protocol approved by the Ethical Committee of Tohoku University School of Medicine and the experiments were performed in compliance with national legislation and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association (Declaration of Helsinki).

Experimental procedure

The experiment comprised three parts: two fMRI sessions and an intervention. Figure 1 summarizes the experimental procedure. Prior to the experiment, the participants were divided into three groups at random, controlling for an equal number of male and female participants in each group. Seventeen participants (13 males and 4 females) were assigned to the first group (DIFF), who were tasked with taking a viewpoint other than their own for the intervention. Seventeen participants (13 males and 4 females) were assigned to the second group, (SAME), who were tasked with taking a viewpoint that aligned with their own perspective for the intervention. Finally, 18 participants (13 males and 5 females) were assigned to a control group (CONT) that did not undergo an intervention.

Figure 1. Outline of the entire fMRI experiment. First, the participant moved to the MRI room and had the first fMRI run (pre-run). Immediately after the first fMRI session, the participant moved to the soundproofed room to carry out the intervention. The participants assigned to the DIFF and SAME groups engaged in a 30-minute intervention task. Participants assigned to the CONT group took a 30-minute break. The intervention task involved making a constructive speech for a contentious issue using a designated point of view. The contrary point of view was required for the DIFF group and the original viewpoint was required for the SAME group. Lastly, the participant moved to the MRI room again and completed the second fMRI run (Post-run).

After finishing the preparation period, which included a questionnaire survey and practicing the experimental task, the participants underwent the first fMRI session, which consisted of one practice run and one actual run (pre-run). Immediately after the first fMRI session, they engaged in a 30-minute intervention task outside the MRI scanner before commencing the second fMRI session. The second fMRI session consisted of one actual run (post-run) and a structural scan of the entire brain.

Experimental stimuli for the CT intervention and fMRI tasks

To enable the participants to consider a point of view other than their own, nine types of conflicting issues were used, six for each fMRI experimental task and three for the topic of intervention. The selection criteria for these issues were based on well-known facts for which an affirmative or opposite viewpoint would be easy for Japanese university students to establish; i.e., the participants could interpret the viewpoint from the context and keywords (a detailed selection procedure is described in Appendix A1-1). Moreover, it was also expected that participants could declare their own viewpoints to those issues. The following six issues were selected for the fMRI task: (1) yutori-style (educational policy to reduce the hours and contents of the curriculum) low-stress education; (2) education that places emphasis on memorization; (3) increasing sales tax; (4) life-prolonging medical care; (5) promotion of nuclear power generation; and (6) smoking. Three conflicting issues during the intervention stage: (7) alcohol overconsumption in university-aged students, (8) gambling, and (9) the transplantation of organs from brain-dead donors. These issues were not used for the fMRI task to avoid potentially confounding mere-exposure effects.

In addition, text that addressed the reasoning behind another person’s viewpoint on the issue [1–6] were prepared for the fMRI task. We prepared 12–20 reasons for each contentious issue. Each reason was prepared so that university students could easily judge the viewpoint of another person. In total, 102 reasons were used (the detailed selection procedure is described in Appendix A1-2). The mean (SD) Japanese character number of the reasoning text was 36 (11) characters.

To confirm the participants’ viewpoint for each conflicting issue, individuals were asked to present their viewpoint and degree of knowledge of each of the nine issues prior to the fMRI measurement. Their subjective viewpoints were scored for each issue and used to assign a designated point of view for the intervention task. The reported degree of knowledge of each issue was used to exclude unfamiliar issues from fMRI data analysis. The participants’ answers to each question were subjected to a five-point nominal scale: (1) strongly agree, (2) agree, (3) disagree, (4) strongly disagree, and (0) impossible to judge because of poor knowledge. The degree of knowledge was also scaled as follows: (1) I have never heard of this issue, (2) I know only the basic terms related to this issue and do not have enough knowledge to form an opinion, (3) I have heard of the issue from TV news and know about this issue to a certain extent, (4) I have previously considered this issue and know this issue well, and (5) I am currently studying this issue and have specialized knowledge.

fMRI task

Figure 2 shows the fMRI task design. Each fMRI run took the form of a rapid event-related design composed of an experimental task to judge another person’s viewpoint on a specific conflicting issue (Judgment) and a control task. In the Judgment task, the task stimulus consisted of two components: the title of the contentious issue [1–6] and corresponding reasoning text. The title of the issue was displayed in the upper left of the screen in yellow font, and the reasoning text was displayed in the center of screen in white font (Figure 2(a)). Participants were instructed to read the reasoning and judge whether the viewpoint of this person was affirmative. To avoid a mere-exposure effect, the reasoning texts were divided into two sets and different sets of reasons were used in each run. The orders of reasons presented in the pre- and post-runs were randomized to exclude an order effect. A control task was also prepared to account for low-order control on visual stimuli and motor response. In the control task, the letters A – F were presented in yellow font in the upper left of the screen and sentences showing possible affirmative or opposing statements (e.g., “I agree with issue A.” in Japanese) were presented in white font in the center of the screen (Figure 2(b)). Four duplicate pro and con statements were prepared for each letter (A–F) giving a total of 48 statements. Within each run, half of all statements were used. The instructions for participants were the same as for the Judgment task; they were asked to read the statement and judge whether the person who wrote the text agreed with this issue.

Figure 2. Details of the fMRI experimental task design. (a) Example of a visual stimulus for the Judgment task. Note that the original stimuli were written in Japanese. In this example, the word in yellow font (Smoking) indicates a contentious issue. The sentence in white font shows the reason why another person (the other) is taking a specific viewpoint on the issue in two to four lines. Participants read the reason and judged whether they approved of the other’s argument. (b) An example of a visual stimulus for the Control task. A simple letter of the alphabet (e.g., issue a) is used instead of the title of the issue. An affirmative or opposing statement is described in the sentence in place of the reason in (a). The details of the stimulus are the same as those in the Judgment task. (c) Timeline of each fMRI run. The duration of each trial was 6 seconds, and the inter-trial interval was set at 1–3 seconds. The order of each Judgment and Control task trial was counterbalanced.

The time allowed for each task was set at 6 seconds, with the task stimulus disappearing immediately when the participant answered, followed by display of a fixation-cross in place of the stimulus. The time between each trial was pseudo-randomly set at 1–3 seconds, with the fixation-cross being displayed during this interval. There were 75 trials in total (51 for the Judgment task and 24 for the control task) for each fMRI run and the order of the trials was assigned pseudo-randomly. Since there was the possibility that some participants did not have sufficient knowledge about some of the contentious issues, large numbers of trials for the Judgment task were compared with the control task. The total scanning time of each complete run was 10 min 30 seconds, including a 20-second rest period before the first trial, and a 10-second rest period after the last trial. For the experimental task, the practice run was used to confirm the degree of comprehension about the tasks. Contentious issues that were not used in the actual run (principle of free competition) were used for the Judgment task of the practice run.

During the fMRI measurement, the participant was placed in a supine position within the MRI scanner. Stimulus presentation and response collection were administered using Presentation software (Neurobehavioral Systems, Inc., Berkeley, CA, USA) that was installed on a personal computer. An LCD projector located outside the scanner projected the stimuli via waveguide onto a translucent screen, which the participants were able to view using a mirror attached to the head coil of the MRI scanner. Behavioral responses were recorded using a fiber-optic response pad (Current Designs Inc., Philadelphia, PA, USA) positioned at the participant’s waist, which the participant could operate comfortably with the right hand. Participants answered the Judgment and control tasks by pressing a response-box button using their right index or middle finger; assignment of the index or middle finger to answer was decided by each participant.

Intervention

Participants assigned to the DIFF and SAME groups were directed to make a constructive speech regarding an issue, with a designated point of view as the intervention. The designated point of view for the writing exercise was decided by a participant’s original viewpoint for the assigned issue, as assessed by the score received by each participant during the subjective questionnaire. The participants in the DIFF group were asked to write a speech using a perspective contrary to their own original viewpoint. For example, a participant opposed to gambling would be asked to write a speech from an affirmative viewpoint, in favor of gambling. The participants in the SAME group were asked to write a speech using the same point-of-view as his/her original viewpoint. For example, a participant opposed to gambling would be asked to write a speech based on a negative attitude to gambling. The contentious issues (7), (8), and (9) were randomly assigned to each participant as speech topics. If the participant did not have enough knowledge (e.g., he/she answered (1) or (2) for the degree of knowledge) for any of the three issues, another issue was given. The participants were then asked to a prepare a speech manuscript using a paper and pen during the allotted 30 minutes in a soundproof room. Participants were instructed to write enough text that would result in a speech of approximately 3 minutes. They were also notified that their audience would be university students and to provide content that would satisfy the educational level of their audience. During the writing task, access to any literature or the Internet was not allowed to ensure that anything written was a result of their own preexisting knowledge. In contrast, the participants assigned to the CONT group were directed to take a 30-minute break in a soundproof room. The use of any electronic device was prohibited in the CONT group to prevent participants from using the Internet to guide their knowledge on the designated issue.

Questionnaire for the stubbornness characteristic

The personality characteristic for stubbornness was measured by the “power to live” questionnaire, which addresses eight personal characteristics that are advantageous for survival under various adverse circumstances (Sugiura, Sato, et al., 2015). Although several existing survey items include stubborn personality as an explanation of concept (e.g., openness to experience in the big-five personality traits or flexibility in the California Psychological Inventory (Gough, 2000)), the explicit relationship with consensus building was not clear. Thus, we selected this questionnaire. Since we were interested in the stubborn personality, the scoring system for stubbornness was selectively used. In this questionnaire, five items are used to obtain the score for the stubbornness factor and each item is measured by a six-point scale.

fMRI data acquisition

All images were acquired using a Philips Achieva 3T MRI scanner (Philips; Amsterdam, Netherlands). The fMRI time-series dataset for the whole-brain scan was acquired using T2*-weighted gradient echo-echo planar imaging (GE-EPI). The parameters of the experiment were as follows: repetition time [TR, 2000 ms]; acquisition time [TA, 2000 ms]; echo time [TE, 30 ms]; flip angle, 80 degrees; 32 slices; field of view [FoV, 192 × 192 mm; 64 × 64 matrix]; slice thickness, 3 mm; slice gap, 0 mm. We collected 315 scans for each pre- and post-run. In total, 630 scans per participant were included in the data acquisition. To acquire a structural whole-brain image at a fine grain, magnetization-prepared rapid-acquisition gradient-echo (MP-RAGE) images were obtained (TR, 6.5 ms; TE, 3 ms; flip angle = 8 degrees; FoV, 240 × 240 mm; 162 slices; voxel dimensions = 1.0 × 1.0 × 1.0 mm).

Data analysis

Regarding behavioral data, the degree of coincidence with common sense interpretation and reaction time for the Judgment task were analyzed for two reasons. The first was to exclude participants with low performance from further fMRI data analysis. The second was to assess whether the task performance itself had changed due to the intervention. The degree of coincidence with common-sense interpretation was defined as when the participant’s judgment for each reasoning coincided with the result of preliminary experiment 2, involving different participants (Appendix A1-2). The degree was calculated by the number of matched responses divided by the total number of Judgment task trials in which participants possessed sufficient knowledge (scores of (3), (4), or (5)) on the subject. In addition, the degree of correct interpretation and reaction time for the control task were also analyzed.

Data preprocessing and statistical analyzes of the fMRI data were performed using Statistical Parametric Mapping (SPM) 12 (Wellcome Trust Center for Neuroimaging, London, UK) running on Matlab R2017a (Mathworks, Natick, MA, USA). All analyses were performed using the Montréal Neurological Institute (MNI) space. As for preprocessing, the effects of head motion across scans were corrected for by realigning all scans to the mean image. The scanning time lag for each slice was adjusted to the timing of the 16th slice, which was obtained at half of the TR. The structural whole-brain image was co-registered with the first EPI image, and the co-registered structural image was spatially normalized to the MNI-T1 template using a segmentation procedure. Accordingly, normalization parameters between the EPI image and the MNI-T1 template were calculated and all EPI images were spatially normalized using the normalization parameters (2 × 2 × 2-mm voxels). Finally, all normalized EPI images were smoothed with a Gaussian filter in a spatial domain (8 mm full-width at half-maximum).

After preprocessing, the data from five participants were excluded due to excessive head movement (head movement exceeding 3 cm or rotation exceeding 3 degrees on any of the pre- and post-runs). Additionally, the data from one participant were excluded because of his low performance for the Judgment task. This exclusion criterion was set to a degree of coincidence with common-sense interpretation of less than 70%. In total, data from 46 participants (35 males and 11 females; mean age 21.2, SD = 1.5) were analyzed, including 15 for the DIFF group (11 males and 4 females), 17 for the SAME group (13 males and 4 females), and 14 for the CONT group (11 males and 3 females).

Data from the fMRI experiments were analyzed using a conventional two-level approach within SPM12 (Friston et al., 1994). The first level assessed the hemodynamic responses produced under the different conditions at each voxel on a subject-by-subject basis using a general linear model. Trials were divided into three conditions to identify the neural mechanisms that were activated when participants interpreted the opinions of the other individuals. Trials in which the correct response was recorded for the Judgment task were grouped into the Judgment condition. Trials in which the participant did not have sufficient knowledge (e.g., he/she answered (1) or (2) for the degree of knowledge) or clear point of view (e.g., he/she answered (0) for the subjective viewpoint) of the issue were not included in the Judgment condition and, instead, were grouped into a condition of no-interest. Trials in which the incorrect response was recorded were also grouped into the no-interest condition. All trials from the control task were grouped into a low-order control condition (Control). To depict the difference of the intervention effect, each condition was modeled separately in the pre- and post-runs. For each participant, Judgment_Pre, Control_Pre, no-interest_Pre, Judgment_Post, Control_Post, and no-interest_Post were modeled. The neural responses for each condition were modeled using a delta-function convolved with a canonical hemodynamic response function (HRF). The occurrence time for each expected activity was set to begin when the participant pressed the button to answer the question. To eliminate any influence of visual processing, neural responses to each visual stimulus were also modeled under the no-interest condition. The response was assumed to be a boxcar function convolved with the canonical HRF. The occurrence time for each neural response was set for when the visual stimulus was presented, and the responses’ temporal span was set for the time until the stimulus was removed from the screen. Additionally, the estimated parameters of head movements obtained by the preprocessing were also included in the design matrix to exclude the effect of head motion. Low-frequency confounding effects were removed using a high-pass filter with a 128-second cutoff. Multiple regression analyses were performed on each voxel to detect the regions in which changes in the MR signal were correlated with the hypothesized model and a partial regression coefficient for each voxel was obtained.

Contrast images obtained through subtraction of parameter estimates were also subjected to second-level analysis. To identify differential activations between the three kinds of interventions, contrast images were created by extracting the differences between the Judgment and control conditions for each run and group. Calculations for this analysis were as follows: (Judgment_Pre > Control_Pre: Pre-DIFF) and (Judgment_Post > Control_Post: Post-DIFF) for the DIFF group; (Judgment_Pre > Control_Pre: Pre-SAME), and (Judgment_Post > Control_Post: Post-SAME) for the SAME group; (Judgment_Pre > Control_Pre: Pre-CONT) and (Judgment_Post > Control_Post: Post-CONT) for the CONT group.

The second level of analysis was performed between groups using a two-way mixed factorial design that analyzed one between-subject factor (group: CONT, SAME, and DIFF) and one within-subject factor (run: pre- and post-run). It should be noted that statistical tests using SPM12 that involve any between-subject factors within a mixed factorial design cannot be valid because of the complexity associated with variance partitioning (Chen, Adleman, Saad, Leibenluft, & Cox, 2014). Thus, the analysis was carried out using GLM Flex Fast 2 software (McLaren, Schultz, Locascio, Sperling, & Atri, 2011). Since the present study has two purposes, we must ensure the independence of the analysis for each purpose (Kriegeskorte, Simmons, Bellgowan, & Baker, 2009). For that reason, only the contrast images from post-run were used to depict the effect of the CT intervention on the neural substrate of open-minded thought. Conversely, only the contrast images from pre-run were used to examine the relationship between stubborn personality and the neural substrate of open-minded thought without the effect of the intervention. To identify changes in neural activation that reflected the participant considering the viewpoint of another, a subtraction analysis between Post-DIFF and Post-SAME (Post-DIFF > Post-SAME) was performed. In this analysis, we hypothesized that there would be no difference between each group during the pre-run, as the participants had yet to experience the intervention in the pre-run phase of the study. Accordingly, the subtraction of the contrasts during the post-run phase was also analyzed. The statistical threshold was set at p < 0.05 and corrected for family-wise error (FWE) by voxel-level inference. The obtained activation peaks were anatomically labeled using the Anatomy toolbox atlas (Eickhoff et al., 2005) within bspmview software (Spunt, 2016).

Region of interest (ROI) analysis was then performed to plot task-specific activation and clarify any correlation between the activation profile within each ROI and the stubbornness index. ROI analyses were performed using the MarsBaR 0.44 toolbox (Brett, Anton, Valabregue, & Poline, 2002) and R version 3.3.3 software (R Core Team, 2017). ROIs were defined from significant activation clusters obtained by subtraction analysis of Post-DIFF > Post-SAME on a second-level. Because activation of the right parieto-frontal network was expected to represent open-minded thought, three activation clusters located on the right orbitofrontal cortex (OFC), dorsolateral prefrontal cortex (DLPFC), and inferior parietal lobule (IPL) were selected as ROIs.

To examine the correlation between the activation profile and stubbornness index, only the activation profiles of (Judgment_Pre > Control_Pre) obtained during the pre-run were used. This was because cognitive processing during the pre-run should not have been influenced by any individual intervention procedure, meaning that the data from the three groups could be pooled. It was expected that task performance (the degree of coincidence with common-sense interpretation for the Judgment task) would become a covariate for each activation profile and the stubbornness score. In addition, sex was included as a control variable but age was not used because the stubbornness index has a significant relation to sex but not to age (Sugiura, Sato, et al., 2015). Partial correlation analyses were performed to compute the residuals of the activation profiles on the pre-run and stubbornness score using task performance and sex as control variables. A permutation test was conducted to assess the statistical significance of correlation coefficients independent of the ROI determination by eliminating the influence of the potential biases of the data; 100,000 permutations were generated by a random sampling without replacement, which preserved the intervention groups. A statistical threshold was set at p < 0.05 (two-sided) and the Bonferroni method was applied to correct for multiple comparisons.

Results

Behavioral results

Table 1 summarizes the average reaction times for each task and run, and the degree of coincidence with common-sense interpretation for the Judgment task and degree of correct interpretation for the control task for each run, respectively. To assess the effectiveness of intervention for each group, changes in the reaction times between the pre- and post-run were also analyzed. The difference in reaction time between the Judgment and Control task for each run was used for this analysis. Two-way ANOVA was performed on groups that differed in their between-subject factor (group: CONT, SAME, and DIFF) and/or their within-subject factor during the pre and post-runs. ANOVA showed no significant main effect or interaction between these groups [main effect of run: F(2,43) = 1.45, p = 0.25, generalized eta² = 0.06; main effect of run: F(1,43) = 1.26, p = 0.27, generalized eta² = 0.003; and group x run interaction: F(2,43) = 0.20; p = 0.82, generalized eta² = 0.001].

Table 1. The average (SD) reaction time (RT) [s] and the degree of coincidence with common-sense interpretation of the Judgment task and the degree of correct interpretation of the control task in the CONT, SAME, and DIFF groups.

	Pre-run		Post-run
Group	Judgment task	Control task	Judgment task	Control task
Average RT [s]
CONT	3.32 (0.61)	1.22 (0.32)	3.30 (0.83)	1.12 (0.34)
SAME	3.26 (0.69)	1.38 (0.54)	3.08 (0.61)	1.12 (0.29)
DIFF	3.01 (0.78)	1.21 (0.30)	2.92 (0.68)	1.11 (0.32)
Degree of coincidence with common-sense interpretation (Judgment task) and degree of correct interpretation (control task)
CONT	0.920 (0.028)	0.985 (0.035)	0.920 (0.045)	0.991 (0.018)
SAME	0.908 (0.069)	0.978 (0.055)	0.929 (0.045)	0.988 (0.020)
DIFF	0.936 (0.036)	0.997 (0.011)	0.920 (0.033)	0.986 (0.026)

A partial correlation analysis for the average reaction time of the pre-run Judgment task and the stubbornness index was also performed. This analysis was used to determine the relationship between a stubborn personality and the processing speed with which the subject was able to infer the other’s opinion. The average reaction time of the Control task on first run, task performance, and sex were used as control variables. Results showed a statistically significant negative correlation between the reaction time of the Judgment task during the pre-run and stubbornness score (partial correlation coefficient = −0.391; t = −2.721; degree of freedom (df) = 41; p = 0.010) indicating that participants with a stubborn personality tended to respond to the Judgment task more rapidly.

fMRI results

The fMRI analysis in this study aimed to elucidate the differences in cortical activation associated with interpreting the opinion of another person with opposing views after experiencing different kinds of intervention. Table 2 summarizes the peak locations of task-related activation that were specific to the DIFF group during the post-run (Post-DIFF > Post-SAME). Significant differences in activation were observed in the right parieto-frontal network of DIFF participants. Two especially active regions were observed on the right lateral side in the OFC and DLPFC, which were anatomically labeled as the middle orbital gyrus and middle frontal gyrus, respectively. A third region with increased activation was located in the right IPL. Significant activation was also observed in the left middle frontal gyrus, right middle temporal gyrus, and right cerebellar posterior lobule (lobule VI). Figure 3 illustrates the activation peak locations (Post-DIFF > Post-SAME) and Figure 4(b–d) shows local activation profiles of the activation clusters in the right parieto-frontal network.

Table 2. Regions showing significantly different activation for the DIFF and SAME groups after experiencing intervention (Post-DIFF > Post-SAME).

Location		MNI coordinate (mm)				cluster size
L/R	Anatomical label	x	y	z	t-value	(voxel)
R	Middle orbital gyrus	38	46	4	5.966	40
R	Middle frontal gyrus	40	34	34	5.742	22
L	Middle frontal gyrus	−38	30	42	5.860	12
R	Inferior parietal lobule	48	−56	52	5.850	15
R	Middle temporal gyrus	50	−26	−6	5.462	1
R	Cerebellum (Lobule VI)	16	−80	−10	6.385	38

The statistical threshold was p < 0.05, corrected for FWE by voxel level. L/R indicates left (L) or right (R) hemisphere. Anatomical labels are based on the Anatomy Toolbox Atlas (Eickhoff et al., 2005).

Figure 3. Results of the fMRI analysis. Regions showing significant activations associated with the Judgment task specific to the DIFF group on Post-run (Post-DIFF > Post-SAME; p < 0.05, corrected for FWE by voxel level inference): (a) right middle orbital gyrus, (b) right middle frontal gyrus, (c) left middle frontal gyrus, (d) right middle temporal gyrus, (e) right inferior parietal lobule, and (f) right cerebellum. Those activations were overlayed on the Colin27 template that was provided by MRIcron software (http://people.cas.sc.edu/rorden/mricron/index.html). The color-bar indicates the t-value for each activation.

Figure 4. Results of the ROI analysis. (a) Definition of ROIs: three activation clusters on the right parieto-frontal network that were obtained by the subtraction analysis of the Post-DIFF > Post-SAME on the second level were used. The color-bar indicates the t-value for each activation. The box-and-whisker plots depict the distribution of parameter estimates of “Judgment – Control” on Post-run for each group (e.g., Post-CONT, Post-SAME, and Post-DIFF, respectively) in (b) the right inferior parietal lobule, (c) the right middle frontal gyrus, and (d) the right middle orbital gyrus. (e) Relationship between the activation profile of the right middle orbital gyrus and the stubbornness score. The dotted line indicates the regression line between them.

Relation of stubborn personality to the task-related activation

Table 3 summarizes the results of the correlation analysis between task-related activation and stubbornness score for each ROI. The activation profile on the right OFC and the stubbornness score showed a statistically significant negative correlation following the removal of sex and task performance (correlation coefficient = −0.370; p = 0.012). Figure 3(e) demonstrates that a relationship between the activation profile on the right OFC and the stubbornness score. Right OFC activation was not involved in participants with a strongly stubborn personality when completing the Judgment task. By contrast, a correlation analysis on the right DLPFC and right IPL did not show a significant partial correlation.

Table 3. Results of partial correlation analysis between the activation profile of each ROI and stubbornness scores.

Anatomical label of ROI(Peak location)	Correlation coefficient	p-value	Statistical significance
R. middle orbital gyrus (x, y, z) = (40, 48, 4)	−0.370	0.012	p < 0.05
R. middle frontal gyrus (x, y, z) = (40, 34, 34)	−0.213	0.155	N.S.
R. interior parietal lobule (x, y, z) = (48, −56, 52)	−0.094	0.530	N.S.

Anatomical labels for each ROI are based on the peak location of the activation shown in Table 1. The correlation coefficient represents the degree of association between the average activation profile of each ROI and the stubbornness index with the effect of sex and task performance removed. Statistical significance was set at 0.05 and corrected for multiple comparisons using the Bonferroni method.

Discussion

The present study examined whether cortical activation associated with the mind-set to interpret another’s opinion would be changed by debate practice in which one argued from another’s perspective. The present study also investigated the relationship between the observed cortical activations and the personality trait of stubbornness. A series of fMRI experiments was conducted in concert with CT intervention, in which the participant made a constructive speech using a designated viewpoint on a contentious issue. In the scanner a task was undertaken in which the participant was asked to infer another’s opposing opinion for several contentious issues. The results show that the right DLPFC, OFC, and IPL were specifically activated after participants constructed a speech using a designated viewpoint that differed from their own when compared with the participants who constructed a speech from their own viewpoint. This result supports the present hypothesis that experiencing another’s perspective enhances activation of the right parieto-frontal network, which was expected to underlie the cognitive components of open-minded thought. In addition, correlation analysis between a stubborn personality and the neural responses associated with interpreting the opinion of another person with opposing views showed that the magnitude of activation of the OFC correlated with a stubborn personality, while the IPL and DLPFC did not show significant correlations. Thus, the results imply that cognitive processing influenced by CT practice is reflected in daily behavioral traits associated with consensus building or discussion.

Differential activation between each group after experiencing the intervention

The present results show that increased activation of the right parieto-frontal network comprises two distinct areas according to the relationship with the stubbornness score: the OFC region showing negative correlation and the IPL-DLPFC network showing no significant correlation. Previous studies have suggested that the right ventrolateral prefrontal cortex (VLPFC) and its neighboring regions serve as an inhibitory pathway for self-perspective (Hartwright et al., 2012; van der Meer et al., 2011). By contrast, DLPFC and parietal activations are involved in updating causal associations during belief revaluation (Corlett et al., 2004, 2006; Fletcher et al., 2001; Turner et al., 2004).

The right OFC region is also associated with pro-social cognitive control, including the inhibition of self-perspective. Previous review articles have supported this notion by pointing to the observation that cortical activation is observed in a broad area of right middle and inferior frontal gyri in association with the selection of appropriate reappraisals (Buhle et al., 2014) as well as self-perspective inhibition. Voxel-based morphometry studies have also suggested that a relationship exists between pro-social processes and OFC function. The gray matter volume of the right OFC or frontopolar regions reflects the size of a subject’s social networks (Lewis, Rezaie, Brown, Roberts, & Dunbar, 2011; Powell, Lewis, Roberts, García-Fiñana, & Dunbar, 2012). The regions reported in those studies were close to our findings in relation to the right OFC region. If the participants’ brains were prepared for open-minded thought by the intervention, they might have spontaneously inferred the other person’s perspective when judging their opinion during the post-run. Therefore, they might also have inhibited their own perspective to interpret the opinions of others fairly. Right OFC activation would also be promoted when pro-social changes of thought occur. Because changes in OFC activation were specifically observed in the DIFF group, the experience of taking the viewpoint of the other did serve as an important preparation step for promoting open-minded thought. The results also show a negative correlation between a stubborn personality and right OFC activation before the intervention. Previous psychological studies have also reported a negative correlation between the stubbornness index and agreeableness in the Big-Five personality traits (Honda, Hirose, & Sugiura, 2018). These findings imply that stubborn personality characteristics may also suppress pro-social cognitive control, reducing right OFC activation.

In contrast, enhancement of right IPL-DLPFC activation could be considered to reflect changes in the cognitive processing of participants when open-minded thought was encouraged following the intervention. Belief revaluation is one of several kinds of causal association learning (Fletcher et al., 2001). High-performance learners showed more activation of the right inferior parietal and middle frontal regions when unsuccessful feedback differed from their own beliefs was received during medical-decision learning (Downar, Bhatt, & Montague, 2011), and activation peaks of reported regions were almost consistent with our activation peaks on IPL-DLPFC regions. A similar activation pattern was also reported in a previous study about cognitive processing of regret. Negative feedback can influence the cognitive processing in subsequent decision making, especially when regret is experienced in the preceding trial. This feedback may be the reason for enhanced choice-related activity in the right IPL-DLPFC regions (Coricelli et al., 2005). The intervention may have compelled the participants to reevaluate both self-belief and contrary opinions to accomplish the constructive speech writing task with a contrary viewpoint. This experience was able to induce a change in mind-set that enabled reevaluation of the association between the contentious issue and another person’s opinion. Thus, activation of the right IPL-DLPFC network was enhanced with the increased cognitive demand for revaluation of an opinion during the intervention.

Taken together, the present results and previous findings indicate that changes in neural responses following the intervention may reflect modulation in the cascades of the right OFC and IPL-DLPFC networks as well as stubborn personality. First, activation of the right IPL-DLPFC network would directly reflect cognitive demand to reevaluate the other person’s opinion. Second, the right OFC would be associated with pro-social cognitive control that drives the reevaluation process needed to consider the opinion of another. Finally, the stubborn personality would be associated with increased upper-layer processing that modulates various kinds of cognitive processing, including the right OFC function. However, cognitive functions promoted by the intervention might not be influenced by the stubborn personality. The behavioral results also show that participants with a higher stubbornness index tended to respond quickly. This result supports a previous report that found a stubborn personality was related to the comprehension speed of reading (Niikuni, Mizuno, Muramoto, & Sugiura, 2018). However, in our study there was no significant difference in the reaction times of the DIFF group between the pre- and post-runs. Therefore, the relationships between a stubborn personality and reaction time, and a stubborn personality and OFC activation, may be independent.

In the present study, a viewpoint contrary to the participant’s own perspective was used as one of the experimental conditions. This was because a negative emotional response, such as antagonism, would be induced more explicitly against the contrary viewpoint and, therefore, the effects of the CT intervention (i.e., taking another’s perspective) could be observed based on the differential activations between intervention groups. The present fMRI results support this prediction. On the other hand, it is also possible to take a viewpoint that is different but still similar to one’s own, as in the case of debate practice. In this case, it was expected that the similar viewpoint would also induce cognitive processing associated with perspective-taking, although that the degree of the negative emotional response might be relatively small. The present study was, however, unable to directly address this point because comparisons between contrary and similar viewpoints would be necessary to clarify the effects of different viewpoints on the change in mind-set.

Implications

The present results showed that changes in the right parieto-frontal activation reflected changes in mind-set due to the CT intervention more clearly than behavioral data such as reaction time. Because the CT intervention, taking another’s perspective, was able to induce open-minded thought, the reaction times of the DIFF group would be expected to become slower. However, there were no statistically significant differences between the DIFF and SAME groups in the Post-run whereas the fMRI results revealed a significant difference between these groups. This discrepancy may be consistent with discussions regarding the necessity of careful consideration when inferring an underlying cognitive process that causes a difference in reaction times (Krajbich, Bartling, Hare, & Fehr, 2015). On the other hand, conventional test measures obviously possess other kinds of advantages, such as being able to investigate a large sample size relatively easily compared with the neuroimaging approach. Moreover, with the neuroimaging approach it would also be difficult to carry out an on-site measurement of the actual situation because of the limitation of the instruments. Thus, it would be possible to evaluate the effectiveness of the CT practice, making good use of both measures complementarily.

Furthermore, the present results imply two-kinds of applicative contributions. First, they show the possibility of constructing a more efficient procedure to promote open-minded thought. Previous studies have reported several beneficial effects from a full set of debate practice tasks (Hall, 2011; Zare & Othman, 2013; Mumtaz & Latif, 2017). However, the full set of debate practice tasks requires many participants (e.g., debaters on both-sides and de-briefers) and enough practice time, including preparation and/or instruction periods. By contrast, the present results demonstrate that open-minded thought can be also induced simply by practicing taking the perspective of another person into consideration. Second, the present procedure might apply a preparatory study of open-minded thought for a constructive discussion outside the educational context. Another advantage of the experimental design was that open-minded thought was promoted by considering different issues to the target issues of the fMRI task. Thus, a practical application of the present procedure could be as a warm-up exercise. For instance, community members could successfully induce open-minded thought by performing an intervention prior to an actual discussion.

Limitations

In terms of the effectiveness of CT practice, the present study utilized only a facet of CT practice because it used only the preparation of constructive speech as the intervention. Debate practice includes other important components such as, cross-examination, rebuttal, and summation processes. Thus, different neural responses might be promoted by practicing other components of debate. Similarly, the present study could not elucidate the improvement of CT skills other than open-minded thought following the intervention. In the intervention period, the participants did not have any technical guidance or practice with regard to writing the constructive speech task. Thus, further investigation using fMRI to characterize neural activity during different kinds of tasks will be necessary to examine changes in skills due to CT practice.

Conclusion

The present CT intervention involving taking the perspectivSe of another person enhanced neural responses in the right parieto-frontal network. Furthermore, the observed neural substrates could be divided into two distinct components. One was the right OFC, which is associated with pro-social cognitive control and exhibited activation that was correlated with a stubborn personality. The other was the right IPL-DLPFC network, which was associated with reevaluation of another’s opinion and which would be reinforced under enhanced pro-social cognitive control. In addition, the present results also imply some applicative contributions, such as development of simple and efficient CT practice, and preparatory work to promote open-minded thought for constructive discussion.

Acknowledgments

The authors thank Y. Akimoto, T. Araki, S. Hanawa, and Y. Kotozaki for their assistance with the preliminary psychological experiments. This work was supported by JSPS KAKENHI Grant Number JP17H06219. Part of this study was carried out under the Cooperative Research Project of the Institute of Development, Aging and Cancer, Tohoku University.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Japan Society for the Promotion of Science [KAKENHI Grant Number JP17H06219].

Appendix

A1: Summary of the preliminary experiment

To determine a suitable experimental setting, two kinds of psychological experiments were conducted before the actual fMRI experiment. All participants in the preliminary experiments provided written informed consent in response to an experimental protocol approved by the Ethical Committee of Tohoku University School of Medicine and the experiments were performed in compliance with national legislation and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association (Declaration of Helsinki).

A1-1: Preliminary experiment 1

The purpose of preliminary experiment 1 was to determine suitable contentious issues for the main fMRI experiment. First, the experimenters selected drinking on the part of university students, gambling, and transplantation of organs from brain-dead donors as topics of intervention because it was thought that these issues were relatively familiar topics for the university students who were the intended participants in the main fMRI experiment. Twelve issues were prepared as candidate fMRI experimental stimuli: advanced maternal age, automobiles, copyright protection, easier education, education that places emphasis on memorization, immigration, increased sales tax, life-prolonging medical care, principle of free competition, promotion of nuclear power generation, simultaneous recruiting of new graduates, and smoking.

Second, several contentious issues were examined as a psychological experiment. Forty-one university students (17 males and 24 females; mean age = 21.0, SD = 2.8) participated, and none of these subjects participated in the fMRI experiment. The participants were asked to choose and rank three of twelve factors, in descending order, of estimated disadvantage for their own life in society. They were also asked to write an essay about each factor, explaining why he/she felt a disadvantage in relation to this issue,within 30 minutes. Six successful candidates were identified by the number of total choices and the content of the written essay. As a result, six contentious issues (easier education, education that places emphasis on memorization, increase in sales tax, life-prolonging medical care, promotion of nuclear power generation, and smoking) were selected as contentious issues for the fMRI experiment.

A1-2: Preliminary experiment 2

The purpose of preliminary experiment 2 was to prepare the experimental stimuli, that is, personal opinions on a contentious issue for both affirmative and opposite viewpoints. To this end, 116 opinions were prepared as candidates for the experimental stimuli. These opinions were based on well-known facts, convenient for people with either a favorable or contrary viewpoint, having been referenced from literature and websites.

To exclude sentences that the intended participants in the main fMRI experiment might find difficult to understand, a psychological experiment was carried out using 11 university students (eight males and three females; mean age = 20.8, SD = 1.1); none of these participants took part in the fMRI experiment. In this experiment, the combination of another person’s viewpoint (affirmative or contrary) about the contentious issue and opinion text for the issue were displayed on the screen of a laptop computer. The participants were instructed to judge whether the subject’s viewpoint and the content of the opinion text were consistent using a numeric keypad. The response time for each sentence was not limited.

The exclusion criterion was that sentences with a proportion of responses in which the consistency of the viewpoint and content of the opinion text was less than 30% were excluded. As a result, 12 sentences were excluded; in addition, one sentence for promotion of nuclear power generation and one sentence for smoking (with the least accuracy) were also discarded due to adjustment of the number of whole sentences. In sum, 102 sentences survived as experimental stimuli: 17 sentences for easier education, 12 for education that places emphasis on memorization, 17 for increased sales tax, 16 for life-prolonging medical care, 20 for promotion of nuclear power generation, and 20 for smoking. The average (SD) proportion of responses that the viewpoint and content of the opinion text were consistent for the selected sentences was 0.78 (0.18), and the average (SD) reaction time for a correct answer was 3.92 (1.06) seconds.