Sensory multi-brain stimulation enhances dyadic cooperative behavior

Abstract

Hyperscanning research suggests that interbrain synchronization supports the regulation of social behavior. However, the evidence is predominantly correlational, leaving a gap for epiphenomenal accounts, where synchrony merely represents concurrent stimulus processing rather than a mechanism relevant to interpersonal interactions. Here, we demonstrate that interbrain synchrony causally drives cooperative success, as evidenced by non-invasive stimulation enhancing coupling and subsequently improving performance in a concurrent interdependent cooperation task. We applied dual-sensory entrainment at 16 Hz and 40 Hz to dyads and compared their performance with non-entrained control dyads performing the same cooperation task. We found that dual stimulation improved interbrain synchrony at the targeted frequencies relative to controls, with 16 Hz entrainment producing the most prominent effect. Strikingly, sensory entrainment facilitated sustained behavioral coupling, allowing partners to maintain coordination over extended periods. Notably, these effects are contingent on improved response coordination, indicating the importance of interbrain coupling for facilitating coordination and demonstrating causally that partner neural attunement is necessary to produce effective joint behavior. Thus, our study supports the concept that interbrain synchrony represents a neural mechanism with functional specificity in social interactions.

Introduction

Cooperative behavior is essential to human social life. It helps with group cohesion, joint decision-making, and effective problem-solving. Recent advances in social neuroscience have created new opportunities to study the brain processes behind cooperation. In particular, the hyperscanning technique allows the simultaneous recording of brain activity of multiple individuals during real-time social interactions (Montague et al., 2002). Using this approach, researchers have consistently reported the emergence of interbrain synchrony (IBS), a phenomenon in which the neural activity of interacting individuals becomes temporally aligned during cooperative tasks (Dumas et al., 2010; Barraza et al., 2020; Czeszumski et al., 2020).

IBS has been observed across most EEG frequency bands, including theta (Lindenberger et al., 2009), alpha (Dumas et al., 2010), beta (Novembre et al., 2017), and gamma (Barraza et al., 2020). These findings suggest that IBS is a neurophysiological marker of social coordination. However, most of the current evidence is still correlational. This point raises a key question: Does IBS cause cooperative behavior, or is it just a result of shared sensory and motor experiences? This debate has recently drawn the attention of several researchers, who argue that the functional significance of IBS will remain speculative without experimental manipulation (Novembre & Iannetti, 2021), explanation (Moreau & Dumas, 2021), and theoretical integration with the broader neuroscience literature (Hamilton, 2021).

Although it is a topic of evident importance, researchers have recently begun to explore the causal influence of IBS on social interactions through externally induced synchrony. For example, nasally administered oxytocin has increased alpha-band IBS during interpersonal coordination (Mu et al., 2016). Electrical dual-brain stimulation with transcranial alternating current stimulation (tACS) enhances movement synchrony and interpersonal learning at 6 Hz (Pan et al., 2021) , improves motor coordination and cooperation at 20 Hz (Novembre et al., 2017; Lu et al., 2023), and sharpens conceptual alignment at 40 Hz (Chen et al., 2022; Liu et al., 2023). Relatedly, transcranial direct current stimulation (tDCS) promotes cooperative strategies in the Prisoner’s Dilemma (Zheng and Lu, 2025). These findings demonstrate that neurochemical and electrical interventions can modulate IBS to boost social interaction. However, whether sensory multi-brain stimulation (MBS) (Novembre & Iannetti, 2021), delivered through synchronized rhythmic visual or auditory input, can similarly shape cooperative behavior remains unknown. By addressing this gap, our study introduces a more naturalistic means of probing interbrain coupling, complementing and extending current electrical MBS evidence.

The present study investigated whether IBS induced through visual sensory entrainment, an experimental technique that uses rhythmic stimulation to modulate neural oscillations at specific frequencies (Ronconi & Melcher, 2017; Lakatos et al., 2019; Albouy et al., 2022), acts as a causal mechanism supporting dyadic cooperative behavior. Based on current evidence, if IBS underlies cooperation, then externally increasing it via dual-sensory entrainment should improve cooperative success compared to a no-entrainment group. Conversely, if IBS is merely an epiphenomenon, no significant differences in cooperative success should emerge between the entrainment and no-entrainment groups. To test this hypothesis, we recorded behavioral responses and EEG activity from adult dyads performing a concurrent interdependent cooperative task (Cui et al., 2012; Liu & Pelowski, 2014), randomly assigned to one of two groups. The experimental group performed the task under four visual entrainment conditions. In the two synchronous conditions, we applied zero-lag dual stimulation to dyads, so both participants received identical flickering sequences with matched frequency and phase. In the two asynchronous conditions, stimulation was delivered at the same frequency but with randomly delayed onsets (5–12 frames, ≈30–73 ms), creating trial-by-trial variability to disrupt in-phase neural alignment. The control group performed the task without entrainment. These manipulations allowed isolation of synchronous and asynchronous entrainment effects on cooperation, controlling for rhythmic stimulation. Visual stimuli flickering at 16 Hz and 40 Hz were applied within the MBS framework to drive interbrain synchrony. These frequencies target bands linked to social coordination: beta for motor alignment (Novembre et al., 2017) and gamma for interpersonal coupling and shared intentionality (Barraza et al., 2020), allowing us to test whether cooperation is affected by entrainment in general or there are frequency-specific effects. IBS was assessed by measuring phase synchronization between pairs of electrodes from the two EEG caps.

Methods

Participants

We estimated the sample size using G*Power 3 (Faul et al., 2007). An effect size (partial eta squared) of 0.117 was obtained from an experimental task similar to that used in this study (Barraza et al., 2020). To achieve a statistical power of 0.95 (α  =  0.05), the minimum sample size was 13 dyads per group for a mixed analysis of variance (ANOVA) with two groups with four measurements each. Therefore, we randomly paired participants from a pool of 60 participants to form 30 same-sex dyads. Dyads were randomized to one of two groups, each consisting of 15 dyads: an experimental group (EXP) (male-male dyads = 7, mean age = 24.2 years, SD = 2.8) that received sensory stimulation while performing the task, or a control group (CON) (male-male dyads = 7, mean age = 27.9 years, SD = 7.1) that did not receive stimulation while performing the same task. Participants within each dyad were unacquainted prior to the experiment. The groups were equivalent in gender and age (t = –1.88, P = 0.077, Cohen’s d = 0.69). All participants were native Spanish speakers, right-handed, with normal or corrected-to-normal vision/hearing, without self-reported history of neurological/psychiatric disorders or drug abuse, and were unaware of the purpose of the experiment. Prior to testing, the participants gave their written informed consent. This study followed the Declaration of Helsinki (1964) and was approved by the Scientific Ethical Committee for the Health Sciences of the Pontifical Catholic University of Chile (CEC-SaludUC).

Task

Experimental design

The experiment involved a concurrent interdependent cooperative task (Cui et al., 2012; Liu & Pelowski, 2014) (Fig. 1A), comprising four conditions of 40 trials each. Each condition was separated by a 30-second rest period. Each trial began with the presentation of a centered fixation cross lasting 2 sec, followed by a cue stimulus consisting of a centered white circle for a random duration between 2 and 2.5 sec. The target stimulus appeared, consisting of a black dot at the cue’s center. Dyad members were asked to press a button on their respective keyboards at the onset of the target stimulus. The cue and target stimuli stayed on the screen until the participants responded. A feedback screen was then presented for 4 sec, indicating whether the trial was a win or a loss and who responded faster or slower (plus and minus signs, respectively). Participants were instructed to work cooperatively to win as much as possible. To win, they had to respond in a coordinated manner such that the difference in their reaction times (RT) did not exceed an adaptive threshold, computed as 1/5 of the average dyad reaction time. This threshold makes the task slightly more complicated than the original proposal (Cui et al., 2012), compensating for possible increases in performance due to stimulus manipulation, resulting in a ceiling effect. The threshold is calculated as follows:

Figure 1.

Task, setting and entrainment effects. (A) Schematic representation of a trial. (B) Participants sit together and perform the joint task without seeing each other. Stimuli are presented in duplicates on each half of the screen. (C) Mean Power Spectral Density changes computed from posterior sites (O1, O2 and Oz) for an exemplary dyad. Entrainment induces power increases at the target frequencies relative to the baseline period for both partners, as depicted by the red asterisks. Red dashed lines represent the robust significance threshold based on two median absolute deviations (MAD) above the PSD median. The same pattern is observed for the 16 Hz Rand and 40 Hz Rand entrainment conditions.

Trial t is a winning trial if

$$\mathit{abs}\left({RT}_{S1,t}-{RT}_{S2,t}\right)<\mathrm{ }\frac{1}{5}m\mathit{ean}({RT}_{S1,t},\mathrm{ }{RT}_{S2,t})$$

where abs() is the absolute value and RT_{Si, t} is the reaction time of subject i during trial t.

For subsequent analyses, we measured each experimental condition’s behavioral performance at the dyad level. For each pair of partners, we calculated the average reaction time (RT) as a basic measure of response speed. We then computed the absolute RT difference between partners to evaluate the closeness of their motor responses. Cooperative success was measured by the percentage of winning trials (PWT). This metric served as our primary index of task performance. To determine the specific behavioral benefits from entrainment, we calculated performance gains. We did this by subtracting the mean PWT of the control group from the PWT of each experimental condition. Finally, to capture the temporal stability of behavioral coordination, we calculated the cooperation streak, which is defined as the mean number of consecutive winning trials.

Sensory entrainment

For the EXP group, we implemented a sensory multi-brain stimulation approach (Novembre & Iannetti, 2021), where cue and target stimuli were presented with a flickering pattern at a fixed frequency of 16 Hz or 40 Hz. For each condition, one of the following patterns was used: (i) a 16 Hz flickering pattern presented synchronously to both participants; (ii) a 40 Hz flickering pattern presented synchronously to both participants; (iii) a 16 Hz flickering pattern presented asynchronously to each participant; (iv) a 40 Hz flickering pattern presented asynchronously to each participant. Synchronous flickering (Sync) was applied to align partners’ EEG activity to the same frequency and timing, facilitating IBS. Conversely, the goal of asynchronous flickering (Rand) was to hinder IBS by randomly delaying the flickering presented to one of the participants. As a result, the frequency was preserved for the dyad, but the timing was disrupted. Stimulus flickering followed a 50% duty cycle square-wave pattern, and flickering operated independently for each participant’s screen half. To achieve asynchronous flickering in the 16 Hz Rand and 40 Hz Rand conditions, the flickering for one participant was randomly delayed by 5–12 frames (approximately 30–73 ms) in each trial. The delay generated a same frequency phase-lagged flickering pattern relative to the other participant, aiming to hinder interbrain coupling. The CON group retained the four-condition structure but did not include any flickering.

Experimental setting

Participants were seated side-by-side, approximately 25 cm apart, facing a shared monitor screen (27-inch; 165 Hz refresh rate) at a distance of 60 (± 3) cm. We used a single, shared monitor to ensure perfect temporal alignment of the flickering visual stimuli for both participants during dual-EEG recording. This setup minimizes synchronization delays and jitter, crucial for avoiding stimulus-driven artifacts in the interbrain signal. An opaque curtain (150 cm high × 90 cm wide) prevented partners from seeing each other. It divided the monitor into two sections, ensuring each participant could see only their respective half of the screen. Stimuli were duplicated on both halves of the screen (Fig. 1B). The participants were instructed to look at the screen while keeping their hands on their respective keyboards. Instructions were provided at the beginning of the experiment. The participants were asked to use earplugs during the experiment to avoid external auditory cues that could influence their performance. Stimuli presentation was controlled using PsychoPy3 v2021.1.4 (Peirce et al., 2019), and behavioral responses were recorded using keyboards.

EEG data acquisition

We followed the EEG hyperscanning setup described by Barraza et al. (2019). The EEG signals of both participants were simultaneously recorded with two 32-channel actiCap electrode caps, each in a separate and ground-isolated BrainAmp DC amplifier (Brain Products GmbH) inside an electromagnetically shielded room. Signals were digitized at 1000 Hz, online filtered between 0.01 and 250 Hz, and synchronously recorded in the BrainVision Recorder software. The electrodes were placed at standard positions based on an extended 10–20 system. Eye movements were monitored using bipolar electrooculogram (EOG) channels. Specifically, horizontal eye movements (HEOG) were acquired by placing one electrode at the lateral canthus of each eye. Vertical eye movements (VEOG) were acquired by placing one electrode below and another above the right eye. The ground electrode was placed at the Fpz location, and the FCz channel was used as an online reference. The electrode impedances were maintained below 10 kΩ.

EEG data preprocessing

BrainVision Analyzer v2.1 (Brain Products, GmbH) was used for offline data preprocessing, and each subject was treated independently. First, continuous EEG signals were filtered between 1 and 45 Hz with a fourth-order zero-phase-shift Butterworth IIR filter. Then, a semi-automatic inspection of the raw data was performed using the built-in BrainVision Analyzer algorithm. The algorithm removed intervals of 200 ms if there were signal values less than 0.5 μV for time windows of 100 ms or if voltage changes exceeded 50 μV/ms, or a difference in values in intervals exceeding 200 μV/ms. Blinks and eye movement artifacts were corrected using the algorithm of Gratton et al. (1983) with VEOG and HEOG channels as templates. The clean EEG signals were re-referenced to an average reference, excluding the VEOG and HEOG channels. Finally, the EEG signal was segmented into epochs lasting 5000 ms, including 2000 ms preceding cue onset.

Interbrain coupling analysis

The epoched EEG data were analyzed using the MNE-Python (Gramfort et al., 2013) and HyPyP Python packages (Ayrolles et al., 2021). First, the EEG signals were Hilbert transformed and decomposed into four frequency bands: theta (3–7 Hz), alpha (8–12 Hz), beta (14–18 Hz), and gamma (38–42 Hz). We centered the beta and gamma bands at stimulation frequencies with a span of ± 2 Hz to match the other two bands. The phase component was used to calculate the phase-locking value (PLV) as a measure of neural synchrony (Lachaux et al., 1999). This measure reflects the inter-trial consistency of the phase difference between all possible pairs of electrode channels, and is computed as follows:

$$\mathit{PLV}\mathrm{ }=\mathrm{ }\frac{1}{N}\left|{\sum }_{n=1}^N{e}^{i\left({\varphi }_j\right(n)\mathrm{ }\times \mathrm{ }{\varphi }_k^{\mathrm{*}}(n\left)\right)}\right|$$

where the vectors Φ_j(n) and Φ_k(n) represent the phases at trial n for the signals at electrode j of subject 1 and electrode k of subject 2, respectively. Φ* denotes the complex conjugate of Φ, and | . | denotes module operator. N corresponds to the number of trials.

The PLV ranges from 0 to 1, with 0 representing the total absence of synchrony (i.e., phase differences are random) and 1 indicating perfect synchronization (ie, the phase differences are constant throughout the trials). For each frequency band, PLV estimates were normalized to a 500 ms baseline (from −1500 ms to −1000 ms) preceding cue stimulus onset and then averaged over 500 ms time windows. This resulted in 10 windows: four on the fixation, five for the cue, and one for the target. The first two windows were discarded prior to statistical analysis.

Statistical analysis

Behavioral data

The behavioral measures were subjected to a mixed repeated-measures ANOVA, with FREQUENCY (16 Hz, 40 Hz) and RELATIVE PHASE (sync, rand) as within-subject factors and GROUP (exp, con) as the between-subject factor. The α level was set at 0.05. Effect size was computed as partial eta-squared values (η_p²). While ANOVA is robust and appropriate, spurious differences might arise from environmental commonalities affecting both participants within a dyad (Holroyd, 2022). We also conducted permutation tests (10⁴ iterations) separately for within- and between-subject effects to account for potential spurious results. Data from all dyads were pooled, regardless of group labeling, and all possible same-sex dyad combinations were constructed. Similarly, condition labeling was shuffled individually for within-subject permutations. Samples were randomly drawn from this pseudo-dyad population and compared, yielding a surrogate test statistic. After iterations, an empirical null distribution can be derived, and a two-tailed P-value computed as the proportion of surrogate statistics whose absolute value exceeds the absolute value of the actual test statistic. Importantly, although permutation tests rule out multiple comparison problems, conducting several permutation tests requires addressing multiple comparisons. Accordingly, all P-values were adjusted using Sidak’s correction for stricter control of Type I error than the more commonly used False Discovery Rate (FDR) approach (Horvath, 2011). This ensures more conservative inferences across dyads and conditions.

Entrainment validation

As entrained brain activity is expected to present a power peak at the stimulation frequency, we analyzed the individual power spectra to confirm the effects of sensory entrainment. For each condition, each participant’s mean Power Spectral Density (PSD) of the posterior sites (channels O1, Oz, and O2) was computed from epoched data, averaged over trials, and normalized to the baseline window described above. The normalized PSD was thresholded using the full spectrum’s median absolute deviation (MAD) such that peaks greater than two MADs above the PSD median were considered significant (Fig. 1C).

Interbrain coupling

We performed cluster-based permutation tests to assess the significance of IBS differences between groups (Maris & Oostenveld, 2007). First, for each condition, individual recordings of both groups were pooled, and the population of all possible same-sex pseudo-dyads was constructed, computing IBS for each pseudo-dyad, as described previously. Second, for each time window and frequency band, samples were randomly drawn from this surrogate dataset and iteratively passed together with the real dataset to MNE-Python, where ANOVAs were performed to compute surrogate F-statistics. Third, these values were used to construct null distributions based on the data clustering. Finally, the real test statistic was compared with this distribution to obtain a P-value. All P-values obtained through this procedure were adjusted using Sidak’s correction.

Results

Sensory entrainment

The entrainment effectively increased the power at the target frequencies, as summarized in Table 1. At the subject level, 76.7% of the EXP participants had significant power increases for all conditions, with 96.7% successfully entrained in both 16 Hz conditions and 80% successfully entrained in both 40 Hz conditions. At the dyad level, this translated to 53.3% EXP dyads with concurrent significant power increases for all conditions, with 93.3% being successfully entrained with both 16 Hz protocols and 60% being successfully entrained with both 40 Hz protocols. Importantly, at least one dyad member was effectively entrained under all the conditions.

Table 1.

Proportion of subjects and dyads showing significant entrainment at the target frequency for each condition

	Overall		16 Hz Sync		16 Hz Rand		40 Hz Sync		40 Hz Rand
Level	n	%	n	%	n	%	n	%	n	%
Subjects (N = 30)	23	76.7	29	96.7	30	100	25	83.3	26	86.7
Dyads (N = 15)	8	53.3	14	93.3	15	100	10	66.7	11	73.3

Interbrain coupling

Entrainment affected the phase of brain activity and thus synchrony. Cluster-based permutation tests confirmed a significant increase in beta-band IBS in the 16 Hz Sync condition compared with the control group. Specifically, this effect was observed from 500 ms to 2000 ms after cue onset (windows 4–6: all P-values < 2 × 10⁻⁴, Sidak corrected) (Fig. 2). No other time window or frequency band was significant for this condition, and no significant differences in IBS were observed for any other condition (all P-values > 2.1 × 10⁻³, corrected).

Figure 2.

Entrainment increased beta-band IBS when subjects were stimulated synchronously at 16 Hz. Significant intersubject links are displayed for the three time windows in which differences were found between EXP and CON groups. Color intensity reflects the number of links connecting each channel.

Cooperation rate

Overall, the cooperation rate was higher in the EXP group (65.7%) than in the CON group (53%), as confirmed by a significant main effect of GROUP (F_(1,28) = 10.38, P = 3.2 × 10⁻³, η_p² = 0.27). Moreover, there was a significant GROUP × FREQUENCY interaction (F_(1,28) = 18.32, P = 2 × 10⁻⁴, η_p² = 0.4, Fig. 3A), with pairwise comparisons showing that only 16 Hz entrainment protocols yielded significant performance increases compared to the control group (P = 6.2 × 10⁻⁵). Within the EXP group, there was a significant effect of FREQUENCY (F_(1,14) = 15.37, P = 1.5 × 10⁻⁴, η_p² = 0.52), which further indicates that 16 Hz entrainment has a greater effect on performance than 40 Hz entrainment. The performance gains are shown in Fig. 3B. No main effects or interactions were associated with the RELATIVE PHASE factor.

Figure 3.

Performance measures of EXP versus CON dyads. (A) Mean cooperation rates for all EXP conditions and the CON group. Dyads cooperate significantly more with 16 Hz entrainment than control. (B) Performance gains in the EXP conditions (average minus mean control). (C) Mean RT for all EXP conditions and the CON group. No differences in speed are observed. (D) Mean differences in closeness of response for all EXP conditions and the CON group. Overall, responses under entrainment are closer than control. (All) Bars show the mean and the standard error. 16 S = 16 Hz Synchronous entrainment, 16 R = 16 Hz Asynchronous entrainment, 40 S = 40 Hz Synchronous entrainment, 40 R = 40 Hz Asynchronous entrainment, ns = nonsignificant.

Reaction times

There was no main effect of GROUP (F_(1,28) = 1.49, P = 0.23) on RTs between the EXP and CON groups, nor a significant interaction between GROUP and the within-group factors (all ps > 0.21, Fig. 3C). Similarly, within the EXP group, neither the main effects nor interactions were significant (all ps > 0.18), suggesting that dyads responded with comparable speed with and without sensory entrainment.

Closeness of response

We assessed the closeness of the responses as a measure of coordination. We found a significant main effect of GROUP (F_(1,28) = 4.26, P = 0.048, η_p  ² = 0.13, Fig. 3D). No other two- or three-way interactions were significant (all ps > 0.13), indicating that partners’ responses are more coordinated under entrainment, regardless of frequency and flickering onset. Similarly, no other main effect or interaction was significant within the EXP group (all ps > 0.75), confirming that responses under any entrainment protocol were similarly close in time.

Cooperation streaks

Next, we wondered if dyads showed a specific behavioral pattern due to entrainment or were randomly cooperating more. To this end, we computed the mean length of the cooperation streaks for each condition and repeated the analyses. Results replicate the trend observed for cooperation rate, with a main effect of GROUP (F_(1,28) = 7.97, P = 8.7 × 10⁻³, η_p  ² = 0.22), and a significant GROUP × FREQUENCY interaction (F_(1,28) = 21.3, P = 7.9 × 10⁻⁵, η_p  ² = 0.43, Fig. 4A), with pairwise comparisons showing that only 16 Hz entrainment yields longer cooperative streaks compared to the control (P = 3.0 × 10⁻⁴). Furthermore, there was a significant effect of FREQUENCY within the EXP group (F_(1,14) = 15.85, P = 1.4 × 10⁻³, η_p  ² = 0.53), indicating that the increase in streak length was higher under 16 Hz entrainment than under 40 Hz entrainment. Again, no main effects or interactions were associated with the RELATIVE PHASE factor. This result supports the notion that dyads engage in behavioral coupling states that allow them to cooperate for extended periods when entrained at 16 Hz. To illustrate this, Fig. 4B shows the trial-to-trial performance of exemplary EXP (upper row) and CON (lower row) dyads for both the 16 Hz Sync and 16 Hz Rand entrainment conditions. While both dyads have similar mean cooperation rates (70% and 67.5%, respectively), it is observable that winning trials of the EXP dyad tend to group. In contrast, winning trials for the CON dyad seem more uniformly distributed.

Figure 4.

(A) Cooperation streak length for all EXP conditions and the CON group. Streaks are significantly longer with 16 Hz entrainment compared to control. Bars show the mean and the standard error. (B) Trial-to-trial performance of an exemplary EXP (upper row) and CON (lower row) dyad for both 16 Hz entrainment conditions. Note that winning trials of the EXP dyad tend to group, while trials for the CON dyad seem more uniformly distributed. 16 S = 16 Hz Synchronous entrainment, 16 R = 16 Hz Asynchronous entrainment, 40 S = 40 Hz Synchronous entrainment, 40 R = 40 Hz Asynchronous entrainment.

Finally, to assess the robustness of our results, we repeated the behavioral analyses using between- and within-group permutation tests (10⁴ iterations). The results confirm that cooperation rates under 16 Hz Sync and 16 Hz Rand entrainment are higher than the control (both ps = 4 × 10⁻⁴, corrected). For cooperation streaks, 16 Hz Sync significantly differed from the control (P = 2 × 10⁻³, corrected), while 16 Hz Rand failed to reach significance (P = 4.4 × 10⁻³, corrected). No differences were found in RT and closeness of response. For within-EXP group comparisons, the main effect of FREQUENCY was marginally significant for cooperation rate (P = 2 × 10⁻³, corrected) and significant for cooperation streaks (P = 1.4 × 10⁻³, corrected).

Discussion

Understanding whether IBS reflects a mechanistic basis for social behavior or is merely a byproduct of shared sensorimotor demands is a central challenge in social neuroscience (Hamilton, 2021; Moreau & Dumas, 2021; Novembre & Iannetti, 2021). Our study addresses this issue by employing rhythmic sensory stimulation to modulate IBS causally during dyadic interactions. We showed that entraining partners’ brain activity leads to significant increases in cooperation and behavioral coordination, thus proving that IBS is a functional mechanism contributing to aligning partners’ cognitive processes.

Crucially, we demonstrated that IBS can be systematically modulated through non-invasive dual visual entrainment, independent of task demands. This methodological contribution meets the criterion of causal manipulation (Herrmann et al., 2016; Novembre & Iannetti, 2021), allowing us to move beyond correlational evidence. Importantly, increased IBS, specifically in the beta band, translated to improved cooperative performance and sustained coordination, supporting the view that neural alignment between brains actively supports joint action (Dumas et al., 2010; Novembre et al., 2017; Barraza et al., 2020). To our knowledge, this is the first study to implement a causal hyperscanning paradigm using sensory MBS. Sensory MBS is different from other non-invasive methods in how it works and its uses. Unlike neurofeedback, which requires active self-regulation, MBS passively aligns interbrain activities through synchronized sensory inputs. In contrast to tDCS or tACS, which apply electric stimulation to brain regions, sensory MBS uses bottom-up sensory pathways to support neural alignment among individuals. This makes it especially effective for improving real-time social interactions and coordination in everyday situations.

Our hypothesis predicted that synchronous entrainment would lead to enhanced cooperation compared to asynchronous or no entrainment. Our results mostly supported this, as entrained dyads showed an overall improvement in cooperation rates, which were significantly higher than those of non-entrained dyads. Entrainment enhanced performance not by improving efficiency, that is, faster RTs, but by improving coordination between participants. Nonetheless, although behavioral coordination increases were comparable between entrainment conditions, they only translated to significant increases in cooperation when stimulation was set at beta frequencies (16 Hz). This aligns with earlier research demonstrating that beta band tACS stimulation facilitates interpersonal motor coordination (Novembre et al., 2017). It also aligns with the hypotheses of beta oscillations as the core neural rhythm subserving action-oriented behavior (Di Dona & Ronconi, 2023) and the maintenance of the sensorimotor status quo (Engel & Fries, 2010). In contrast, entrainment at gamma frequencies (40 Hz) did not yield statistically significant improvements in cooperation, although we did observe a trend towards increased cooperative performance. This result could be attributed to the inconsistency of 40 Hz entrainment on brain activity at the dyad level, as stimulation was only effective for approximately 70% of the dyads. Current research shows that while visual entrainment is effective up to approximately 100 Hz (Herrmann, 2001), the response to flickering decreases as the frequency increases (Zhigalov et al., 2019). Furthermore, the brain slightly habituates to 40 Hz entrainment over time (Petro et al., 2024). Thus, our stimulation may have been less effective in driving gamma-band IBS and not achieved sufficient strength to produce reliable results. This could partly explain its marginal effect on cooperation compared to 16 Hz stimulation. Future research exploring the role of gamma-band IBS may benefit from optimized protocols for entraining gamma oscillations (Lee et al., 2021; Petro et al., 2024).

We found no significant differences in cooperation rates between the synchronous and asynchronous entrainment conditions. This finding should be considered carefully. Previous studies have shown that only in-phase stimulation improves behavioral performance (Novembre et al., 2017; Pan et al., 2021). In our design, asynchronous flickering included a random lag for one participant that stayed the same within each trial. This set up a stable lag-phase relationship, producing a form of dual entrainment in which participants were phase-locked but not strictly in-phase, nonetheless enhancing cooperative performance. This helps explain why we did not see the expected behavioral differences. Future studies could achieve better control over entrainment by randomizing the flickering phase within the trials. Alternatively, modulating the stimulation frequency for participants around the frequency of their partner could also be helpful.

Additionally, our findings reveal an important and often overlooked aspect of interbrain coupling dynamics. The absence of differences in the cooperation rate between our synchronous and asynchronous entrainment conditions suggests that social behavior may be robust to constant phase delays of interbrain coupling. Since daily social interactions are not perfectly synchronous, this form of coupling might better resemble the natural mechanisms for coordination in the brain. Consequently, it would be interesting to investigate whether more precise social action coordination is associated with a finer interbrain phase coupling between interacting partners.

Finally, cooperation rates were consistently higher with 16 Hz than with 40 Hz entrainment. The cognitive processes associated with these frequency bands can explain the superior impact of beta frequencies. The decisive factor for cooperative performance in our task is behavioral coordination, ascribed to IBS in the beta band (Novembre et al., 2017), rather than psychological or affective alignment associated with gamma-band IBS (Barraza et al., 2020). Beta-band entrainment promoted the emergence of behavioral coupling states between partners, sustained over multiple trials rather than transient coincident responses. Dyads showing increased beta-band IBS coordinated better and maintained cooperation across trials. This aligns with the status quo hypothesis of beta oscillations (Engel & Fries, 2010) and extends to joint actions in social interactions. Partners had to calibrate and sustain motor responses over time, relying on coupled beta oscillations to succeed. Sustaining such states requires analyzing fluctuations to predict a partner’s behavior, consistent with anticipatory processes described by the status quo hypothesis and the mutual prediction theory of social behavior (Kingsbury et al., 2019; Hamilton, 2021). Supporting this, previous research linked beta-band IBS to anticipation (Dumas et al., 2010), and dyadic co-modulation of beta oscillations was associated with mutual prediction of actions (Rosso et al., 2022). Thus, beta-band IBS facilitates cooperative behavior, demonstrating functional specificity (Dumas et al., 2010) and causal prominence (Ross & Bassett, 2024a; Ross & Bassett, 2024b; Tseng & Cheng, 2024) for interbrain coupling in social coordinations.

This study has some limitations. First, the small sample size could have led to overestimated effect sizes and unreliable statistical outcomes. Indeed, all our comparisons yielded large effect sizes. However, this could reflect the strength of our stimulation protocols, and our power analysis confirmed an adequate statistical power. Larger replication studies will help to confirm the robustness of our findings. Second, handling dyad interdependence in hyperscanning data remains challenging. One way to account for interdependence has been to include dyad ID as a random factor in linear mixed-effects models when an individual contributes across multiple dyads (Pan et al., 2021). However, no consensus exists for controlling contextual interdependence, commonalities in behavior arising from performing a task together (Holroyd, 2022). Systematic comparisons of approaches, including dyad shuffling, dyad permutation, and split-dyad permutation, would help establish best statistical practices in the field. Third, we must be clear about what we mean by ‘cooperation’ in our research. From basic action coordination to strategic decision-making, cooperative behavior takes many forms. Nonetheless, all of them entail a joint effort to achieve shared goals, grounded in shared intentionality and the capacity to think in terms of ‘we’ rather than ‘I’ (Tomasello, 2009). In this framework, our task demonstrates real-time joint action coordination, in which two partners align their responses dynamically instead of strategically (e.g. Prisoner’s Dilemma task). Future research should examine whether these results apply to other types of cooperative behaviors. Fourth, it remains possible that 16 Hz stimulation enhanced individual-level processes (e.g. response speed) rather than interpersonal dynamics. However, we found that reaction times were unaffected by entrainment. At the same time, cooperation gains emerged as longer streaks of successful trials and tighter temporal alignment between partners, suggesting that the effects are more consistent with a causal role of interbrain coupling than with individual-level factors. To address this, future work should include specific control conditions (e.g. single-participant stimulation).

Conclusion

Our results show that rhythmic dual-sensory stimulation can improve interbrain synchrony. This increase leads to better and lasting cooperative behavior between pairs. These findings support the idea that interbrain synchrony is key in social coordination. They also highlight the specific functions of oscillatory dynamics in social interactions. By showing that brains can be externally aligned to foster cooperation, this study contributes to a growing understanding of social cognition as a dynamic, interpersonal, and oscillatory process.

Author contribution

Ivo Leiva Cisterna (Conceptualization [Lead], Data Curation [Lead], Formal Analysis [Lead], Funding Acquisition [Equal], Investigation [Lead], Methodology [Lead], Project Administration [Lead], Resources [Equal], Software [Lead], Supervision [Lead], Validation [Lead], Visualization [Lead], Writing Original Draft [Lead], Writing Review Editing [Lead]), Paulo Barraza (Conceptualization [Equal], FundingAcquisition [Equal], Methodology [Equal], Resources [Equal], Supervision [Equal], Visualization [Equal], WritingReviewEditing [Equal]), Eugenio Rodríguez (Conceptualization [Equal], Methodology [Equal], Supervision[Equal], Writing Review Editing [Equal]), Guillaume Dumas (Conceptualization [Supporting], Funding Acquisition [Lead], Methodology [Equal], Resources [Equal], Software [Supporting], Supervision [Supporting], Writing Review Editing [Supporting]

Conflict of interest

The authors declare no competing interests.

Funding

IL was supported by ANID Doctoral fellowship (21202583), PB was supported by PIA-CONICYT Basal Funds for Centers of Excellence (FB0003), ANID/Support 2024 (AFB240004), and FONDEQUIP (150003). GD was supported by the Institute for Data Valorization, Montreal (IVADO; CF00137433), Fonds de recherche du Québec (FRQ; 285289), the Natural Sciences and Engineering Research Council of Canada (NSERC; DGECR-2023-00089), the Canadian Institute for Health Research (CIHR 192031; SCALE), and the Azrieli Global Scholars Fellowship from the Canadian Institute for Advanced Research (CIFAR) in the Brain, Mind & Consciousness program.

Data availability

The data underlying this article will be shared upon reasonable request to the corresponding author.