The studies reviewed share similar participant demographics but employ distinct experimental paradigms to investigate interpersonal dynamics (Table 2; Figure 3). Coordination and imitation were explored through finger-tapping (Khalil et al., 2022; Hu et al., 2022) and finger-tracking exercises (Gumilar et al., 2021), while storytelling interactions examined native and non-native speech comprehension (Li et al., 2021, 2023). In the native context, both speaker and listener were Chinese natives, whereas in the non-native study, Chinese speakers were paired with Korean listeners proficient in Chinese. Chabin et al. (2022) simulated a RW concert experience, while Pan et al. (2023) engaged participants in collaborative design tasks. In contrast to prior studies that primarily focused on dyadic interactions, Chabin et al. (2022) included 15 participants, whereas Pan et al. (2023) formed groups of three but recorded neural activity from only two members in each group.

Study designs and environmental conditions varied (Table 2). Finger-tapping (Khalil et al., 2022; Hu et al., 2022) and storytelling studies (Li et al., 2021, 2023) used block designs with rest periods, testing variables such as music and background noise levels, respectively. In contrast, Gumilar et al. (2021), Chabin et al. (2022), and Pan et al. (2023) all used within-subject designs, with Gumilar et al. (2021) and Pan et al. (2023) comparing VR and real-life conditions, while Chabin et al. (2022) focused on the effects of interpersonal distances and music in a concert setting.

Four studies reviewed employed fNIRS (Li et al., 2021; Hu et al., 2022; Li et al., 2023; Pan et al., 2023), while the other three utilized EEG (Chabin et al., 2022; Gumilar et al., 2021; Khalil et al., 2022; Table 3). All fNIRS studies utilized the optimal source-detector distance of 30 mm, allowing for the interrogation of the cortical areas of the brain, except for Pan et al. (2023) who did not specify the source-detector distance used.

Most studies recorded brain activity simultaneously from all individuals within a dyad, or from all 15 participants during the concert scenario (Chabin et al., 2022; Gumilar et al., 2021; Hu et al., 2022; Khalil et al., 2022; Pan et al., 2023). However, two studies investigating the impact of background noise, took a different approach (Li et al., 2021, 2023). In these instances, ‘each dyad consisted of a speaker and a listener. The speaker verbally narrated a story, which was recorded, while their brain activity was monitored. These recordings were then manipulated, by adding various levels of white noise to create four signal-to-noise ratio (SNR) conditions. Subsequently, the recorded audios were played back to the listener while their brain activity was recorded.

The integration of additional physiological metrics was uncommon among the studies surveyed. Only one study reported taking measures such as heart rate and skin conductance, using the external device Shimmer^® Sensing GSR + (Chabin et al., 2022).

All investigations highlighted the frontal lobe’s role in decision-making, emotional regulation, motor planning, and social cognition, with most studies also focusing on the parietal lobe for spatial awareness and sensorimotor integration, and several examining the occipital and temporal lobes for visual processing, auditory processing, and language comprehension (Table 3).

Multichannel configurations play a crucial role in capturing the spatial distribution and dynamics of neural signals across different brain regions. Higher channel density provides finer spatial resolution but may also increase complexity and data processing requirements. All studies reviewed opted to use multichannel probe designs (Table 3). EEG studies used between 5 and 16 channels (Chabin et al., 2022; Gumilar et al., 2021; Khalil et al., 2022), while the fNIRS studies used between 22 and 36 channels (Li et al., 2021; Hu et al., 2022; Li et al., 2023; Pan et al., 2023). It is typical for EEG to employ fewer channels as the scalp’s conductive properties enable EEG signals to be detected by electrodes without requiring a high channel density. The EEG studies also used higher sampling frequencies (128–4,000 Hz) compared to the fNIRS studies (10–17 Hz). This is because EEG directly measures electrical activity in the brain, which fluctuates rapidly. On the other hand, fNIRS measures hemodynamic responses, which change more slowly, allowing for lower sampling frequencies.

Ensuring consistent positioning of optodes over targeted brain regions across subjects is essential to guarantee reliable and comparable data, as slight variations in placement can lead to differences in the brain regions being measured. To achieve this, all studies, except for one (Khalil et al., 2022), utilized the 10–20 system for electrode placement (Jasper, 1958; Okamoto and Dan, 2005). The 10–20 system maps out specific anatomical landmarks on the scalp based on the percentages of the distance between four key reference points: the nasion, inion, and the left and right preauricular points. In addition to employing the 10–20 reference system, 3 fNIRS studies specified further procedures to ensure accurate localization of cortical areas (Li et al., 2021; Hu et al., 2022; Li et al., 2023). One method involved the projection of topographic data onto a 3D reference frame based on skull landmarks, thereby allowing a probabilistic reference to cortical regions underlying each channel (Hu et al., 2022). Other studies employed virtual registration methods to establish correspondence between fNIRS channels and measurement points on the cerebral cortex (Li et al., 2021, 2023). By utilizing a 3D digitizer for precise channel positioning, researchers obtained potential coordinates for each channel. Subsequently, brain regions associated with the channels were determined using the Automated Anatomical Labeling (AAL) atlas (Tzourio-Mazoyer et al., 2002).

In reviewing the preprocessing methods applied across the studies, several commonalities and distinctions emerged. While all studies prioritized the removal of motion artifacts to enhance the quality and reliability of the data, the specific techniques employed varied (Table 4).

Gumilar et al. (2021) employed independent component analysis (ICA) to address multiple artifact types simultaneously, an approach that performs particularly well in the non-Gaussian signal environments typical of EEG data. However, ICA is computationally intensive and generally more effective with larger datasets, so the relatively modest dataset size in this study (16 channels) may have limited its effectiveness. In contrast, Chabin et al. (2022) applied the more computationally efficient principal component analysis (PCA), which removes components explaining the most variance, usually artifacts. PCA’s faster processing makes it better suited to smaller datasets, such as the 14 channels used, though it is more prone to overcorrection and struggles to differentiate between artifacts and meaningful neural signals as effectively as ICA.

Building on the advantages of PCA, targeted principal component analysis (tPCA) was applied by Li et al. (2021, 2023) alongside cubic spline interpolation. This adaptation of PCA is designed to more precisely isolate and remove motion artifacts by focusing on predefined artifact periods, reducing the risk of overcorrection seen in traditional PCA. Given both studies’ relatively high number of channels (36), tPCA was particularly well-suited, allowing for effective artifact reduction without excessive data loss. However, like PCA, tPCA relies on the accurate identification of artifact epochs, which can introduce bias if artifacts are misclassified. Additionally, tPCA can be time-consuming due to the manual selection of parameters.

In contrast, Hu et al. (2022) utilized a correlation-based signal improvement method, which targets the preservation of hemoglobin signal changes by addressing motion-induced correlations between HbO2 and Hb. While this method is effective for neurotypical populations, it is less suited to populations with atypical brain physiology, as it assumes a consistent negative correlation between HbO2 and Hb, limiting its generalizability to more diverse populations (Cui et al., 2010; Obrig and Steinbrink, 2011). Given Hu et al.’s use of more than 20 channels, this method allowed for a robust analysis of hemoglobin signal changes across a larger dataset.

Manual artifact removal, as used by Khalil et al. (2022), allows for human oversight in distinguishing artifacts from neural signals. This method is particularly useful when artifacts are difficult to identify through automated processes. However, manual selection is subjective, time-consuming, and difficult to replicate consistently across datasets and researchers. Given the small scale of this study (only five channels) manual selection was a more appropriate choice, as using ICA or PCA would likely have been ineffective with such limited data.

Filtering was another key preprocessing step employed by four studies (Pan et al., 2023; Chabin et al., 2022; Gumilar et al., 2021; Khalil et al., 2022). In fNIRS, filters are essential for removing physiological artifacts, such as cardiac pulsations and respiratory signals, without altering the functional hemodynamic response. Similarly, in EEG, filters help eliminate low-frequency drift, muscle artifacts, and electrical noise to ensure cleaner neural signals. The choice of filters varied among the studies, depending on the specific noise characteristics and research focus. A detailed review of the advantages and limitations of various filtering techniques is available in Dans et al. (2021).

Both Chabin et al. (2022) and Khalil et al. (2022) also employed data alignment procedures to ensure temporal synchronization across participants. This step is crucial for group-level analyses, particularly in complex neuroimaging studies where even minor misalignments can distort time-sensitive neural response interpretations.

All reviewed fNIRS studies employed Wavelet Transform Coherence (WTC) analysis to measure IBC (Li et al., 2021; Hu et al., 2022; Li et al., 2023; Pan et al., 2023) (Table 5). WTC decomposes time-varying signals into different frequency bands, calculating coherence between the corresponding frequency components of signals from different brains. This method provides an estimate of the degree of synchrony across participants’ brain signals under various experimental conditions. However, while WTC is widely used in fNIRS research (Hakim et al., 2023), it poses certain challenges. One of the main limitations is the ambiguity around the selection of frequency bands in fNIRS, as the physiological meaning of different frequency components is less clear compared to more direct neural measures like EEG. fNIRS signals often include a combination of physiological processes unrelated to cognitive activity, making it difficult to directly associate frequency bands with brain function. Current methods for selecting frequency bands, whether through visual inspection, task-related associations, or significance testing, are often lacking in mechanistic justification. The studies could have strengthened their analyses by pre-registering frequency band selections or providing a stronger theoretical basis for their chosen bands.

In contrast, two EEG studies employed phase-based methods to measure IBC, with each method tailored to the unique demands of the experimental task. Khalil et al. (2022) used Phase Correlation, a technique measuring the alignment between the phases of EEG signals in near-real time. This method was particularly suited for the fast-paced, dynamic nature of their anti-phase tapping game. Phase Correlation’s sensitivity to these moment-to-moment shifts made it the ideal choice for capturing immediate neural alignment, although it is less capable of measuring stable, sustained synchrony over longer periods, which might be important for post-task analyses. On the other hand, Gumilar et al. (2021) employed Phase Locking Value (PLV), which is better suited for measuring the consistency of phase alignment over multiple trials or longer periods. Their finger-pointing and tracking exercise involved repetitive, predictable actions, and PLV’s focus on long-term phase stability allowed the researchers to capture how consistently participants’ brain waves remained in sync throughout the task. While PLV is ideal for structured, repetitive interactions, it is less sensitive to transient changes in synchrony, which could be important in more dynamic or spontaneous tasks.

Chabin et al. (2022) took a different approach by using Total Interdependence (TI) and Theta Coherence (ThetaCo) to analyze IBC during a live concert setting. TI was computed to assess overall cerebral coupling between participants, accounting for feedback beyond concurrently recorded data points, which provided a more comprehensive measure of synchrony compared to traditional coherence. However, TI’s assumption of Gaussian stationarity may be problematic in dynamic settings like a concert, where brain signals are likely non-stationary. Additionally, the study used ThetaCo to focus specifically on inter-brain coherence in the theta frequency band (4–8 Hz), which is often associated with emotional processing. This allowed the researchers to capture neural synchrony related to shared emotional experiences, such as moments of high pleasure during the concert.

In addition to IBC, several studies also conducted single brain analyses to explore intra-brain activations. Li et al. (2021, 2023) analyzed the activity patterns within individual brains to understand how different regions interacted during task and rest periods, using statistical techniques like ANOVA, cluster-based permutation tests, and Pearson’s correlation. Gumilar et al. (2021) extended their analysis by incorporating EEG source localization, which mapped neural activity onto specific brain regions, providing spatial insight into the brain areas involved in IBC.

IBC, as observed across studies, appears to be influenced by changes in the environment such as exposure to music, variations in interpersonal distance, VR simulations, and background noise levels (Figure 4). Moreover, implicated brain regions associated with IBC span various cortical areas (Figure 5; Table 6).

Gumilar et al. (2021) observed that while IBC can be achieved in VR environments, the degree of synchrony may differ compared to traditional face-to-face interactions, suggesting that visual perspective plays a crucial role in modulating neural coherence. In the real world (RW) scenario, IBC significantly increased from pre-to post-training in theta and beta bands. Notably, significant synchrony occurred in cortical regions such as the dorsolateral prefrontal cortex (DLPFC), the primary motor cortex (PMC), the middle temporal gyrus (MTG), the inferior frontal gyrus (IFG), and the somatosensory association cortex (SAC). Contrastingly, in VR, IBC significantly increased post-training in the alpha and beta bands, with additional delta-band activity in the first-person perspective condition. However, these effects were observed in only about 30% of regions of interest (ROIs). The results suggest that while there may be differences in the magnitude or patterns of synchrony, the fundamental mechanism of neural coupling remains intact across both virtual and real-life contexts.

Pan et al. (2023) extended these findings by investigating the effects of VR on IBC during collaborative design tasks. In the idea generation phase, participants in the RW setting demonstrated higher levels of IBC in areas such as the primary motor cortex (PMC) and supplementary motor area (SMA) compared to those in VR, likely due to the RW’s more immersive and tactile nature, better supporting motor and sensory integration. However, during the idea selection phase, VR participants exhibited greater IBC in the prefrontal cortex (PFC) and supramarginal gyrus (SMG). Increased activation in the PFC suggests the virtual environment may enhance cognitive synchronization and facilitate more efficient communication and decision-making. Additionally, VR fostered higher IBC in the left temporoparietal junction (l-TPJ), which is associated with creativity and idea evaluation, indicating that VR may support creative thinking processes. During the idea deepening phase, the VR environment led to greater IBC in the primary somatosensory cortex (S1), possibly due to the larger virtual space and enhanced tools that provided participants with a more stimulating environment for perception and engagement. On the other hand, the RW group displayed higher IBC in the superior temporal gyrus (STG), which plays a key role in social cognition and facial recognition, suggesting that RW interactions better facilitate implicit communication.

The results of the Chabin et al. (2022) study revealed a negative correlation between the proximity of participants and cerebral coherence indices, particularly evident when participants simultaneously reported high pleasure emotions when listening to music. This suggests that closer physical proximity fosters greater IBC implying a form of interindividual emotional influence. Interestingly, the study found no significant relationship between physiological coupling and proximity, suggesting distinct mechanisms underlying cerebral and physiological coherence. The findings also suggest that emotional resonance mechanisms may underlie the observed cerebral coupling, wherein shared emotional experiences lead to mutual oscillatory influences between individuals.

Khalil et al. (2022) investigated the effect of musical interaction on IBC and performance. They found a significant increase in delta band IBC after a musical interaction, particularly at electrodes over the central and right frontal brain areas. Notably, no significant theta-band IBC changes were reported. This suggests that musical interaction can influence brain dynamics at the interindividual level. Moreover, the duration of musical interaction was significantly correlated with delta-mediated IBC in the post-interaction condition, indicating a positive relationship between the duration of musical interaction and the degree of IBC during the interaction. In agreement with this, Hu et al. (2022) results suggest that exposure to a musical meter enhances IBC. IBC in the left middle frontal cortex (MFC), was significantly greater in the meter condition than in the no-meter condition. Moreover, IBC in the left MFC was higher in the presence of strong meters relative to weak meters. These results were further supported in the concert setting, as IBC was higher when people were listening to music compared with the baseline period (Chabin et al., 2022).

Li et al. (2021, 2023) found that background noise modulates IBC in distinct ways across cortical regions. In Li et al. (2021), IBC in left-lateralized language areas, specifically the IFG, Superior Frontal Gyrus (SFG), Middle Frontal Gyrus (MFG) and SMG, decreased as noise increased, whereas right MTG and AG showed stable IBC across noise levels. In contrast, Li et al. (2023) reported increased IBC in right-lateralized auditory-motor regions (r-STG, r-Postcentral Gyrus (postCG), and r-MTG) under the highest noise condition (−9 dB), though this increase was not monotonic across all noise levels. These findings suggest that while background noise can suppress coupling in traditional language areas, it may enhance compensatory coupling in auditory-motor networks under challenging listening conditions.

A noteworthy observation is the heterogeneous approaches adopted by researchers to present their findings on how environmental factors impact IBC and the associated brain regions (Figures 4, 5). This diversity in data visualization methods, ranging from bar graphs and scatter plots to box plots and detailed brain heat maps, reflects not only the varied nature of the investigations but also the challenges inherent in conveying this complex data. Such variability may underscore the necessity for standardized reporting formats in future research to facilitate comparison, synthesis, and meta-analytical assessments of findings across studies. Additionally, among the studies reviewed, only Pan et al. (2023) systematically reported effect sizes (Cohen’s d) across both behavioral and neural metrics, with medium to large effects (Cohen’s d = 0.55–1.63). The remaining studies generally reported statistical significance (p-values) but did not include standardized effect sizes, further limiting cross-study comparisons of magnitude.

Behavior and emotions also exhibited sensitivity to environmental factors such as music, distance, background noise and whether the environment was real or virtual. Closer physical proximity during a concert setting fostered greater emotional sharing among participants, suggesting a link between spatial distance and emotional resonance (Chabin et al., 2022). This was accompanied by increased IBC during high-emotion musical segments, indicating neural synchrony may underpin shared affective experiences. Moreover, the presence of a musical meter improved coordination during the finger-tapping task (Hu et al., 2022). Specifically, the mean interpersonal time lag was shorter in the meter condition. This behavioral improvement was positively correlated with higher IBC, suggesting that neural synchrony facilitates precise joint action. Conversely, Li et al. (2021, 2023) highlighted the detrimental impact of background noise on speech comprehension performance, with higher SNR levels resulting in lower comprehension performance. Interestingly, correlation analyses between speech comprehension performance and neural coupling suggested distinct behavioral relevance patterns for left and right-lateralized clusters (Li et al., 2021). In Li et al. (2021), stronger IBC in the l–IFG predicted better performance under high-noise conditions, while couplings in the right MTG and AG were more predictive of performance at mild noise levels. Similarly, Li et al. (2023) found higher IBC in the r-STG, r-postCG, and r-MTG was associated with better comprehension in noisy settings, highlighting compensatory auditory-motor coupling under degraded input. Finally, collaborative design behavior in VR and RW environments exhibited both similarities and differences (Pan et al., 2023). Although the overall quality of ideas generated by participants in VR and reality did not differ greatly, distinct patterns emerged at different stages of the design process. In the ideation phase, participants in VR tended to produce ideas of lower quality compared to those working in reality. However, they experienced a higher sense of fluency and subjective satisfaction. Additionally, during the elaboration phase, participants in VR expressed higher levels of satisfaction with their individual and team efforts compared to their RW counterparts. Interviews with participants suggested that VR tools fostered a more productive brainstorming process by generating more ideas than expected, despite some concerns about their quality. Notably, during the idea selection phase, participants in VR demonstrated greater originality, potentially due to the ease of communication and collaboration in a virtual setting. This was supported by positive correlations between originality and IBC in the l–PFC for the VR group, and in the l–TPJ and PMC for the RW group. Additionally, fluency in the RW group was negatively correlated with IBC in the l-TPJ, suggesting differing cognitive dynamics across settings.

A summary of the RoB assessment is presented in Table 7. Overall, all included studies were judged to have some concerns regarding RoB, primarily due to limited demographic diversity (e.g., college student samples), small sample sizes, and a lack of analytic techniques to disentangle true interpersonal neural coupling from common input or task-evoked synchrony. This is discussed further in sections 4.2.2 and 4.2.3.

The limited number of papers published highlights the relatively early stage of this field, likely due to challenges associated with experimental protocols, the only recent feasibility of such studies, and funding limitations. Nevertheless, this review underscores the critical importance of considering environmental factors in the study of social interactions and the need for broader exploration.

The findings presented open the door to a new genre of experiments aimed at advancing our understanding of brain-environment interactions and their role in shaping social behavior. Environmental features such as lighting, spatial layout, green spaces and climate may all impact IBC and, by extension, interpersonal dynamics. These elements, both individually and collectively, warrant investigation in both controlled experimental designs and ecologically valid settings. Additionally, investigating coherence across physiological and behavioral markers, such as synchrony in pupil diameter, heart rate alignment, or breathing rate, could complement neural findings and provide valuable insights into shared states and their relationship to social cohesion.

Beyond foundational science, this research holds translational potential across clinical and applied settings. For example, examining how environmental modifications affect social engagement in populations with social-skill deficits, such as individuals with ASD or schizophrenia, may inform low-cost, non-invasive strategies to enhance social engagement. Interventions could include using music during interactions, adjusting interpersonal distance, or reducing sensory load. Additionally, VR platforms, through shared visual perspectives or structured joint tasks, may offer promising tools for social skills training in clinical or developmental populations. In mental health, educational, and workplace contexts, insights from this research can inform practical design choices, including spatial layout, acoustic conditions, and communication tools, to improve interpersonal connection and coordination across settings.

This line of research aligns with a forward-looking vision of social neuroscience that emphasizes studying brain function in RW contexts. By moving beyond traditional, controlled laboratory settings and integrating neural, physiological, behavioral, and environmental measures, researchers can better capture the complexity of human experiences. This shift toward a more integrative and holistic approach will not only deepen scientific understanding but may also hold practical implications for designing interventions and environments that promote mental health and positive social interactions in everyday life.