In a congested situation, a pedestrian needs to determine one’s own pace by interacting with and predicting other pedestrians’ states to avoid collisions. This issue has practical significance in evacuation during emergencies or in designing large social events (Helbing and Johansson, 2009). Auditory rhythm, such as metronome sound or music, is known to have a major influence on human walking (Styns et al., 2007). A previous study reported that pedestrian flow decreased during group walking in a congested situation, whereas the flow improved by playing a repeated sound from a metronome (70 beats per minute; BPM) to pedestrians (Yanagisawa et al., 2012). The fact indicates that the steady beat is a controllable factor that contributes success/failure of coordinated group walking, and suggests a possibility that the steady beat sound affects the pedestrians’ coordination-related cognitive processes for considering the state of other pedestrians in a congested situation. However, no studies have investigated the effect of a steady beat on such cognitive processes in pedestrians.

Inter-personal interactions synchronize inter-personal brain activity in the prefrontal region, which has a role in social interaction (Scholkmann et al., 2013). NIRS-based “hyperscanning” techniques have gained interest in inter-subject neural synchrony (INS) by measuring simultaneously multiple brains that socially interact with each other (Konvalinka and Roepstorff, 2012; Schilbach et al., 2013). A pioneering study using hyperscanning reported that INS between two subjects’ functional near-infrared spectroscopic (fNIRS) signals from the prefrontal cortex increased during a cooperative task in which the subjects simultaneously pressed their buttons (Funane et al., 2011); in contrast, INS did not increase during a competitive task in which the subjects pressed their buttons with the goal of responding faster than their partner (Cui et al., 2012). Furthermore, INS reportedly increased when two subjects more successfully interacted in the button-press task (Funane et al., 2011; Cui et al., 2012). Similarly, degree of relative successfulness of cooperative button-pressing by two subjects was reported to be associated with an increase/decrease in INS (Cheng et al., 2015). From these findings, brain-to-brain coupling is assumed to be caused by successful modeling, which is a learning process by observing behavior of the others, and prediction of the mental or physical states of peers during social interaction tasks. The button-press task shed light on the understanding of INS in the prefrontal region but the task setting is close to a laboratory experiment. Extending the previous findings into more real-life oriented social interaction, several studies showed connection between verbal communication and INS (Jiang et al., 2012, 2015; Nozawa et al., 2016). On the other hand, a previous study showed relationship between physical coordination (i.e., non-verbal interaction) and INS (Holper et al., 2012). The task used in the previous study was a dyadic coordination in which a subject imitated finger-tapping movements performed by the other, and thus the target to be coordinated was specific and focal. However, relationship between collective physical coordination, which is frequent in real-life and requires a more diffuse coordination with the surrounding, and INS remains unclear.

Group walking focused in the present study is a collective social interaction task as pedestrians are required to determine their own pace of walking on the basis of others’ states such as walking pace and physical distance between self and others. In addition, group walking is coordination with the surrounding crowd, where recognition of more than direct neighbors makes difference (Helbing and Molnar, 1995; Suma et al., 2012). Firstly, we confirm the previous result (Yanagisawa et al., 2012) that a steady beat sound (70 BPM) improved walking flow. Secondly and more essentially, we hypothesize that in congested group walking, a steady beat makes INS of the pedestrians’ prefrontal activities higher than that without the steady beat, reflecting the effect of steady beat in driving pedestrians into more successful modeling and prediction of each other’s state. Furthermore, we expect that the effect of the steady beat on the INS could be observed over a spatially extended range, reflecting the diffuse nature of the coordination where coordination with more than direct neighbors is expected to be an important factor for a successful group walking.

Frontopolar region, a part of the prefrontal cortex, was focused by the present study because the frontopolar region is known to be associated with social interaction (McCabe et al., 2001; Sänger et al., 2011). The frontopolar region showed higher activity during cooperative button-pressing compared to other prefrontal parts (Funane et al., 2011). Furthermore, significant INS related to the button-press task was observed in the frontopolar region; INS derived from the frontopolar region tended to increase with increasing task performance (Cheng et al., 2015).

INS may be susceptible to spurious relationships between the frontopolar activities and the external stimulus (i.e., the steady beat) or the resultant behavioral synchronization. To control such spurious relationships, we added stepping on a same spot, which is similar in movement to walking but does not require the subjects to consider others’ states, either with or without the steady beat. Therefore, our experiment consisted of four conditions: sound/no-sound walking and sound/no-sound stepping. We expected stepping did not show a significant difference of INS between presence and absence of the steady beat sound. To measure frontopolar activity, we used ultra-small NIRS (Figure 1; Nozawa et al., 2016) which has many advantages: multi-brain measurement, wireless, no limits for natural movement, and measurement of skin blood flow. To estimate INS, we used wavelet transform coherence (WTC) (Torrence and Compo, 1998; Grinsted et al., 2004). The obtained INS was statistically compared across two factors comprised of the four conditions (1st factor: walking/stepping, 2nd factor: sound/no-sound), using two-way repeated-measures analysis of variance (ANOVA). Expectation supporting our hypothesis was that INS during sound walking would become significantly higher than during no-sound walking, reflecting an interaction effect between the steady beat and group walking.

In our study, ultra-small NIRS was used to measure frontopolar activities while the subjects performed group walking that is a collective social interaction, or group stepping. When a metronome sound of 70 BPM was played during group walking, the walking flow Q significantly increased compared with that during the no-sound condition. The steady beat sound also increased synchronization of stepping between the subjects. For INS, we observed significant interaction effects of the two factors (walking/stepping and sound/no-sound) at the period from approximately 25–26 s, indicating that the beat made INS while walking higher than that without the beat. This effect may reflect the individuals’ improved predictability of the collective physical coordination facilitated by the steady beat sound. Music that includes a steady beat or some beats is commonly used in difficult and complex group actions (e.g., dance). Therefore, our findings suggest a possible effect of music on frontopolar INS also during such group actions. In contrast, no significant interaction effects were observed in skin blood flow, supporting the cerebral origin of the observed interaction effects. Furthermore, we observed the effect of the beat on the INS over a spatially extended range, suggesting the diffuse nature of the coordination. Our study is the first to investigate relationship between collective physical coordination and INS, and to reveal the effect of the beat on INS.

The reason for introducing the stepping conditions was to verify whether the external auditory stimulus and resultant behavioral synchronization induce brain synchrony between the subjects, independently of coordination-related components. If the external stimulus induces brain synchrony between the subjects, INS during stepping with the beat is supposed to be higher than that during stepping without the beat. However, we did not observe such effect of the beat in stepping conditions. Therefore, the significant difference of INS between walking conditions (with and without the beat) is not attributable to the external stimulus.

The previous studies suggested that the coordination-related components of the task contributed to the increase of INS (Funane et al., 2011; Cui et al., 2012). This suggestion led to a question of whether INS for the walking conditions became higher than that for the stepping conditions. To investigate it, we confirmed simple main effects between the four conditions (SW, NW, SS, and NS) on the brain coherence at the significant period ∼25 s, using ANOVA (Tables 1, 2). Accordingly, INS for NW was lower than that for NS, and INSs between SW and SS did not show the significant difference. These results were inconsistent with the suggestion by the previous studies. A likely explanation for the phenomenon is that relative success/failure in coordination is another key factor in the increase/decrease of INS, in addition to the mere existence of demand for coordination in the task. The previous studies used coordinated social interaction tasks (e.g., simultaneous button press) in which subjects could easily cooperate with each other without support to help coordination (e.g., the steady beat). Unfortunately, such easy tasks may not dissociate the importance of the two factors. Our results indicate that even in a situation which involves high demand for coordination, relative failure in coordination (as in NW) could be accompanied by presumably high but uncoordinated recruitment of social processes, which can lead to even lower INS compared to the situations with no coordination demand (as in SS and NS). This interpretation can also reconcile the contrasting result obtained by Holper et al. (2012). They compared INS during a dyadic imitative coordination task between self-paced and stimulus-paced modes, and found higher INS during the self-paced mode compared to the stimulus-paced mode. This may have been because the self-paced mode imposed higher demand for coordination than the stimulus-paced mode, while the relative successfulness in coordination was not significantly different between the two modes (as indicated by “the mean inter-response duration” as their performance measure of coordination).

Furthermore, due to the diffuse nature of the collective coordination in group walking, we expect that the steady beat sound could affect the INS between pairs of pedestrians who are not spatially adjacent. The results of the analysis comparing INS between near/far pairs (Figure 5, Table 3, and Supplementary Figure S3) revealed that the presence or absence of the steady beat sound induced larger difference in INS between far pedestrian pairs rather than near pairs. This is consistent with our expectation, suggesting the possibility that the steady beat sound could facilitate virtual interaction between subjects who are distant from each other, which could improve collective pedestrian flow in crowded situations.

Frontopolar region is known to have a role in social interaction. We confirmed this by comparing the activation level of the measured frontopolar neural signals between the walking and stepping blocks. As a result, we observed significantly higher activation during walking than stepping (Supplementary Figure S4). The result indicated that the measured frontopolar neural signals indeed reflected functions involved in social interaction and coordination, of which demand should be higher during group walking. Therefore, we would speculate that the observed difference in INS induced by the presence/absence of steady beat during walking reflects temporally well-/ill-aligned recruitment of coordination-related cognitive processes, such as mentalizing and predicting each other’s states.

Our study, using ultra-small NIRS, provided a novel experimental paradigm that enabled multi-person studies, i.e., involving more than two subjects, of realistic social interactions in which the subjects had no limits for natural movement. Most previous studies on social interactions have focused on interactive tasks wherein the subjects had some constraints on their motions (NIRS: Cui et al., 2012; Jiang et al., 2012; Cheng et al., 2015; Duan et al., 2015; EEG: Lindenberger et al., 2009; Astolfi et al., 2010; Dumas et al., 2010; Yun et al., 2012) because the signal transfer was restricted by NIRS fibers or electric cords of the EEG. In an attempt to relieve such constraints, using a portable EEG, a previous study showed the possibility of measuring brain activities of multiple persons during interaction wherein the subjects had none of the above-mentioned limits (Gevins et al., 2012). However, EEG measurement is very susceptible to artifacts arising from muscle and eye movements; therefore, NIRS measurements may be considered more reliable for measuring brain activities under conditions having no movement constraints. Two previous studies used wireless NIRS, which overcomes the aforementioned limitations for EEG measurement during social interaction tasks (Dommer et al., 2012; Holper et al., 2012). However, these previous studies did not investigate unconstrained bodily social interactions between >2 subjects. A previous study investigated INS by verbal communication among three subjects using common NIRS with limits for subjects’ movement (Jiang et al., 2015). Our previous work showed the effectiveness of our ultra-small NIRS on studying INS by natural and unstructured verbal communication between four subjects (Nozawa et al., 2016). The present study, furthermore, extends the experimental paradigm of fNIRS hyperscanning to a next step that enables multi-person studies without limits for natural movement, i.e., group walking.

We observed the significant interaction effects in the period from approximately 25–26 s (i.e., 0.038–0.04 Hz). The low-frequency range (0.01–0.25 Hz) could be subdivided into four bands, as previously defined (Penttonen and Buzsáki, 2003; Buzsáki and Draguhn, 2004; Zuo et al., 2010): slow-5 (0.01–0.027 Hz), slow-4 (0.027–0.073 Hz), slow-3 (0.073–0.198 Hz), and slow-2 (0.198–0.25 Hz). Furthermore, gray matter-related hemodynamic oscillations have been reported to primarily occur in slow-4 and slow-5 (0.01–0.073 Hz) (Zuo et al., 2010). Because our significant results were observed between approximately 0.038–0.04 Hz within slow-4, the results were considered to have originated from the gray matter oscillations.

During the resting state, the low-frequency oscillation amplitudes in slow-5 are known to be dominant within medial prefrontal cortices (including part of the frontal pole) relative to that of slow-4 (Zuo et al., 2010). Unfortunately, the frequency dependency of cerebral functions remains unclear. Our results suggested that the oscillations within slow-4 in the frontal pole had an important role in a social interaction task (i.e., group walking). Future research is needed to investigate the functional significance of slow-4 and -5 in various social interactions.

Our NIRS system has limitations on the number of sensors and the coverage, i.e., the frontal pole only. Nevertheless, our findings of relationship between the coordinated group walking, steady beat sound, and INS obtained using the system are novel. Furthermore, the frontal regions, including the frontal pole, have been assumed to be important for social interactions [i.e., the frontal pole (McCabe et al., 2001; Cheng et al., 2015); the right superior frontal cortex (Cui et al., 2012); the left prefrontal region (Dommer et al., 2012), and left inferior frontal cortex (Jiang et al., 2012)]. Therefore, we suggest that our NIRS covering the frontal pole may contribute to better understanding of the neural mechanisms of social cognition. On the other hand, larger number of sensors and wider coverage may bring additional useful information because brain activities related to social interaction are widely distributed in the brain, such as in the temporal regions (Lindenberger et al., 2009) and right parietal region (Dumas et al., 2010).

The present study using NIRS-based hyperscanning revealed that a metronome sound improved pedestrian flow and INS, and that the steady beat sound has an effect of extending range where pedestrians physically coordinate with other pedestrians. The obtained results promise to extend understanding of relationship between physical coordination and INS.

SI, TN, and RK designed the research. SI, TN, RY, AM, YS, and KS performed the experiments. SI analyzed the data. SI and TN wrote the manuscript. All authors reviewed the manuscript.

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.