Inducing Interpersonal Closeness: Effects on Inter‐Brain Synchrony and Collaborative Performance

ABSTRACT

Feeling close to another person is widely assumed to facilitate collaboration. However, causal evidence that interpersonal closeness improves collaborative behavior remains limited, and it is unclear whether any such influence is accompanied by increased synchrony at the neural level. Previous studies comparing inter‐brain synchrony (IBS) and collaborative performance across relationship types (e.g., friends vs. strangers) have reported better collaboration and higher IBS in closer relationships, but these effects are confounded by familiarity and shared interaction history. To address these limitations, we experimentally manipulated interpersonal closeness in 123 dyads of strangers (N = 246) to test its causal influence on IBS and collaborative performance. Dyads were randomly assigned to one of three conditions: No Interaction, Small Talk, or a closeness‐generating conversation. After the manipulation, dyads completed a collaborative task while undergoing mobile EEG hyperscanning. The manipulation was effective: perceived closeness increased significantly between the No‐Interaction, Small‐Talk, and closeness‐generating conversation conditions. However, induced closeness did not affect collaborative performance or IBS during collaboration. These findings suggest that the better collaboration and higher IBS often reported for pre‐existing relationships in previous studies may originate from factors beyond momentary feelings of closeness, such as familiarity derived from prior interaction. Interestingly, across all dyads, we observed a negative association between central delta IBS and collaborative performance, with IBS explaining unique variance beyond self‐reported measures. This negative link is consistent with the idea that, in this context, IBS reflects speech‐tracking and increased cognitive demands during coordination, rather than coordination efficiency.

Impact

Prior work has linked closer relationships to higher inter‐brain synchrony (IBS) and better collaboration, but these effects may reflect familiarity rather than closeness per se. By experimentally manipulating perceived closeness in 123 stranger dyads, we isolated closeness from familiarity and found that induced closeness neither improved collaborative performance nor increased IBS. Contrary to previous findings, central delta IBS was negatively related to performance, suggesting IBS may reflect speech‐tracking and heightened cognitive demands rather than coordination efficiency.

1. Introduction

Humans are a social species, and collaboration is central to maintaining collective well‐being and societal functioning. To fully characterize this complex behavior, it is necessary to identify the social factors that influence collaboration and to understand the neural processes that support it. A first step in doing so is to adopt a clear definition. According to Dillenbourg (1999), collaboration occurs when individuals work jointly toward a shared goal, hold comparable status, possess similar levels of knowledge, and can perform similar actions. Importantly, collaboration must be distinguished from cooperation, in which partners divide a task into sub‐tasks, work on them independently, and combine their outputs to form the final result. In contrast, in collaborative contexts, partners work on the same problem at the same time, exchanging information through speech, gestures, or other modalities that enable them to influence one another's cognitive processes (Dillenbourg 1999; Roschelle and Teasley 1995).

A key component of collaboration is positive interdependence (Deutsch 1949; Salomon and Globerson 1989; Thibaut and Kelley 1959), defined as the mutual dependence among partners to achieve a shared goal (Johnson and Johnson 2005). Interdependence relies on balance in knowledge, skills, motivation, and effort (Dillenbourg 1999). When this balance is disrupted, one partner may become dominant, leading to free‐riding or social loafing. Such dynamics are especially common in tasks where success depends on the most capable member (Salomon and Globerson 1989). Interdependence is also associated with costs, including increased cognitive effort, potential interpersonal conflict (Johnson and Johnson 2005; Thibaut and Kelley 1959), and in some contexts, stress (Sariñana‐González et al. 2017).

Despite these challenges, collaborative behavior has been conserved across species because of its adaptive advantage: groups often outperform individuals acting alone, and joint action increases the rewards attainable by all partners involved, enhancing overall fitness (Tomasello et al. 2012). The adaptive value of collaboration is further reflected in its early emergence during development. Children display collaborative behavior by 2 years of age in peer dyads (Brownell et al. 2006) and as early as 18 months in interactions with adults (Warneken et al. 2006). These early forms of collaboration are supported by developing socio‐cognitive capacities that allow infants and young children to share intentions and pursue joint goals (Brownell et al. 2006).

1.1. Affiliative Feelings and Collaborative Performance

Collaboration not only maximizes collective gains but also strengthens social bonds, increasing trust, liking, closeness, and similarity in children (Plötner et al. 2015; Yu et al. 2022) and enhancing trust, liking, and closeness among adults (Depping and Mandryk 2017; Johnson et al. 2014; Patané et al. 2020). Moreover, collaborative experiences have been associated with reductions in implicit intergroup biases in adults (Patané et al. 2020) and explicit group biases in children (Yu et al. 2022). In adolescence, collaborative classroom goal structures predict more positive peer relationships than individualistic or competitive structures (Roseth et al. 2008).

Taken together, these findings suggest that collaborative interaction can foster positive social attitudes and affiliative feelings. However, the reverse relationship, whether affiliative feelings, and particularly interpersonal closeness, enhance collaboration, remains unclear, as existing findings are mixed and correlational. For example, Reinero et al. (2021) found no link between group affiliation (liking and feelings of connectedness toward the group) and group performance, contradicting the conclusions of an earlier meta‐analysis (Mullen and Copper, 1994). Other studies have compared collaborative performance across dyads and groups with different pre‐existing relationships, such as friends, romantic partners, or strangers (Brennan and Enns 2015; Djalovski et al. 2021; Jehn and Shah 1997; Pan et al. 2017), and reported that individuals in closer relationships collaborate more effectively. A meta‐analysis by Chung et al. (2018) similarly showed that friends outperform acquaintances. More recent work, however, has found no difference in performance between friends and strangers (Zhang et al. 2024). Importantly, in all of these studies, interpersonal closeness is confounded with familiarity and history of prior interaction, and no causal relationship can be established.

Close relationships typically develop through repeated interactions, during which familiarity, shared history, and affiliative feelings become intertwined. Nevertheless, familiarity and interpersonal closeness are conceptually distinct and may have separable effects on collaborative performance. Repeated contact does not necessarily produce closeness, even when it increases familiarity. For instance, greater familiarity may not be accompanied by stronger affiliative feelings when individuals develop negative impressions of one another or when their interactions remain superficial. The distinction also operates in the opposite direction: perceived interpersonal closeness can develop rapidly without prior familiarity. For example, individuals who share an intense emotional experience may temporarily feel close despite knowing little about one another. Because familiarity requires repeated contact, whereas perceived closeness may develop rapidly, distinguishing between their effects has important practical implications. Specifically, if closeness itself were shown to enhance collaborative performance, it could represent a promising target for interventions in contexts where increasing familiarity is not feasible.

1.2. Interpersonal Closeness and Inter‐Brain Synchrony Between Collaborators

Interpersonal closeness may shape collaboration not only at the behavioral level but also at the neural level. Specifically, it may modulate the coupling of collaborators' brain activity, known as inter‐brain synchrony (IBS; Dumas et al. 2010). IBS is measured using hyperscanning (simultaneous recording of neural activity from multiple individuals), which, thanks to advances in mobile neuroimaging, can now be conducted during real‐time, naturalistic social interactions (Grootjans et al. 2024; Janssen et al. 2021).

Hyperscanning studies of collaborative tasks have shown that stronger IBS is associated with better dyadic and group performance (Cui et al. 2012; Pan et al. 2017; Reinero et al. 2021; Wikström et al. 2022), suggesting that IBS may reflect processes supporting interpersonal coordination and communication. Consistent with this interpretation, IBS has been linked to key components of collaborative interaction, including motor synchrony (Dumas et al. 2010; Koul et al. 2023; Y. Li et al. 2024), engagement (Rai et al. 2025), as well as predictive processes underlying speech and action (Dikker et al. 2014; Z. Li et al. 2024; Mayo and Shamay‐Tsoory 2024; Pérez et al. 2017).

Beyond its association with behavioral coordination, IBS during collaboration may also vary as a function of the social and affective relationship between interacting individuals. In particular, previous studies have reported that dyads in closer relationships not only perform better but also show stronger IBS during collaboration (Djalovski et al. 2021; Pan et al. 2017; Zhang et al. 2024). Yet, similar to the association between interpersonal closeness and collaborative performance, the association between closeness and IBS remains correlational since comparisons between pre‐existing and newly formed dyads are inherently confounded by differences in familiarity and prior collaboration history.

Although several processes may contribute to the observed association between relationship closeness and IBS, two explanations are particularly well supported by the existing literature. First, the feeling of interpersonal closeness itself could influence IBS by promoting spontaneous motor mimicry. Affiliative goals and liking facilitate the mimicry of facial expressions and body movements (Bourgeois and Hess 2008; Chartrand and Lakin 2013; McIntosh 2006), potentially increasing lagged motor synchrony as interaction partners respond to one another's movements. Greater motor synchrony could, in turn, contribute to stronger IBS. This account is consistent with evidence linking IBS to spontaneous motor mimicry during face‐to‐face interaction, including the imitation of body movements, facial expressions, and eye movements (Koul et al. 2023).

Alternatively, familiarity could influence IBS independently of closeness. Through repeated interactions, individuals may become better able to predict a partner's behavior (Cacioppo et al. 2017). Given the association between predictive processes and IBS described above, this enhanced predictability could contribute to stronger IBS regardless of how close partners feel to one another.

In summary, although interpersonal closeness, collaborative performance, and IBS have been linked in previous research, causal evidence for the influence of interpersonal closeness on IBS and collaborative performance is still lacking. Moreover, it is unclear whether any effect of closeness on IBS persists after accounting for spontaneous motor synchrony.

1.3. The Present Study

The present study addresses these gaps by experimentally manipulating interpersonal closeness between dyads of strangers and examining its effects on IBS and collaborative performance. To this end, we used a between‐subjects design in which unacquainted dyads were randomly assigned to one of three conditions: No Interaction (NI), Small Talk (ST), or Fast Friends (FF; Aron et al. 1997). In the NI condition, partners silently reflected on their answers to a set of superficial questions while a screen between them prevented visual contact. In the two interaction conditions (ST and FF), partners engaged in a structured conversation in which they took turns asking one another pre‐defined questions. In the ST condition, questions remained superficial (i.e., identical to those used in the NI condition), whereas in the FF condition, questions became increasingly personal. According to prior work, dyads in the FF condition reliably report stronger affiliative feelings such as closeness, liking, and perceived similarity compared to dyads in the ST condition (Aron et al. 1997; Sprecher 2021). This makes the FF an ideal paradigm for investigating how variations in interpersonal closeness causally impact IBS and collaboration, without the confounds present in pre‐existing relationships. For the same reason, we verified before the study that all paired participants were unfamiliar with each other.

After the manipulation, partners completed a computer‐mediated collaborative maze task (Mills 2014), while EEG and video recordings were obtained to quantify IBS and motor synchrony within dyads.

The three preregistered main objectives of this study were to assess (1) the causal impact of interpersonal closeness on collaborative performance, (2) the causal impact of interpersonal closeness on IBS during collaboration, and (3) whether IBS during collaboration predicts collaborative performance. In addition, the preregistered secondary aims were to test: (1) whether interpersonal closeness influences motor synchrony during collaboration and, if so, whether the effect of closeness on IBS persists when controlling for motor synchrony; and (2) whether collaboration itself increases affiliative outcomes, namely interpersonal closeness, liking, and perceived similarity, and whether this increase is moderated by the type of previous interaction. The study was preregistered on the Open Science Framework (OSF).

Additional exploratory analyses that were not preregistered examined the correlation between central delta IBS and motor synchrony, the correlation between motor synchrony and performance, the predictive value of IBS for performance measured as the number of moves, and changes in IBS between Block 1 and Block 2 of the collaborative game.

1.3.1. Main Preregistered Hypotheses

1. Based on prior research comparing collaboration between relationship types (Brennan and Enns 2015; Chung et al. 2018; Djalovski et al. 2021; Pan et al. 2017), we hypothesized that collaborative performance would be highest in the FF condition, followed by the ST condition, and lowest in the NI condition.

2. In line with previous studies comparing inter‐brain synchrony (IBS) between relationship types during collaboration (Djalovski et al. 2021; Pan et al. 2017; Zhang et al. 2024), we further hypothesized that IBS would be highest in the FF condition, intermediate in the ST condition, and lowest in the NI condition.

3. Based on prior research reporting a positive association between IBS and collaborative performance (Cui et al. 2012; Pan et al. 2017; Reinero et al. 2021; Wikström et al. 2022), we expected that IBS during collaboration would positively predict collaborative performance, replicating earlier findings.

2. Methods

2.1. Participants

The sample included 123 dyads of participants aged 18–35 years (M = 20.8, SD = 2.8), all enrolled in or graduates of tertiary education (Bachelor's, Master's, PhD). Age did not differ across conditions, F(2, 242) = 0.645, p = 0.526. Dyads were gender‐matched (female–female: 96; male–male: 27), reflecting the predominantly female psychology student population at our university (80%). Same‐gender dyads were formed to reduce variance associated with dyad gender composition, which simplifies the interpretation of between‐condition comparisons. Where possible, partners were also matched by first language (Dutch native speakers: 61 dyads; English‐speaking/international: 62 dyads). Condition assignment was stratified to balance the distribution of female–female dyads and Dutch‐native dyads across conditions. Participants with epilepsy or other neurological disorders were excluded. To minimize familiarity effects, friends were excluded, although superficial acquaintances (e.g., classmates) were permitted; baseline familiarity did not differ across conditions, F(2, 120) = 0.276, p = 0.759. Screening for prior unfamiliarity was essential, as the closeness manipulation was validated with strangers (Aron et al. 1997; Sprecher 2021). Additional demographic details can be found in Supporting Information A1.

2.2. Sample Size

An a priori power analysis in G*Power (v3.9.1.7) was conducted to estimate the sample size needed to test the effect of the closeness manipulation on IBS. Modeled as a mixed‐design ANOVA with Condition (NI, ST, FF) as a between‐subjects factor and Region (Frontal, Central, Posterior) as a within‐subjects factor, the analysis indicated that 108 dyads (N = 216; 36 per condition) would be required to detect a medium effect (f = 0.25, α = 0.05, power = 0.80). Because no prior study had tested the effect of a closeness manipulation on IBS, we based this estimate on Sprecher (2021), who reported a relatively large effect (f = 0.40) on perceived closeness in a similar paradigm, but we expected smaller effects on IBS. To allow for potential data loss, we increased the target sample to 120 dyads (40 per condition, total N = 240).

2.3. Experimental Procedure

Participants signed up online and completed a Qualtrics screening questionnaire (consent, eligibility, demographics). All participants provided informed consent to participate in the study. Separate informed consent to publish a blurred photograph of the experimental session was obtained from participants depicted in Figures 1 and 2.

FIGURE 1.

Frames taken from two video recordings showing the experimental setup during the manipulation in (A) interactive conditions (Small Talk and Fast Friends) and (B) No‐Interaction condition.

FIGURE 2.

Frame taken from a video recording showing the experimental setup during the collaborative task. The setup was identical across all experimental conditions.

Dyads were matched by availability, gender, and, when both were Dutch, by first language. To screen for prior acquaintance, participants were emailed their partner's first and last name and asked whether they knew the person; dyads reporting friendship were excluded, whereas superficial acquaintances were retained (e.g., students attending the same course). Before arrival, participants also completed additional online questionnaires testing empathy, social reward, social anxiety, personality traits, and autistic traits (not analyzed here; see preregistration). Upon arrival, two experimenters fitted portable EEG headsets (~20 min) while a room divider prevented interaction between participants. EEG sessions lasted around 2 h and included a 24‐min interpersonal closeness manipulation followed by a collaborative game.

The collaborative game required participants to solve a set of 10 mazes, divided into two blocks of five, separated by a 2‐min break. No time limit was imposed on participants to solve the mazes; therefore, the duration of the game varied between dyads (M = 19 min, SD = 6 min). Before the manipulation, participants rated their prior familiarity with their partner on a sliding scale from 0 (“strangers”) to 100 (“friends”), with 50 indicating “acquaintances.” Most dyads reported low prior familiarity, with only four dyads scoring above 20 (range: 27.5–54.5). Given the low overall familiarity scores, all dyads were retained in the analyses. Interpersonal closeness (the main independent variable in this study), as well as two other affiliative feelings (liking and perceived similarity), were measured before and after the manipulation, and again after the collaborative game (Supporting Information A2). This allowed us to test whether the manipulation successfully induced differences in interpersonal closeness between conditions and whether collaboration itself enhanced these feelings within dyads. After the collaborative game, participants completed a brief evaluation of the collaboration (see Perceived Collaboration Quality).

The present study focuses exclusively on data from the collaborative task; analyses of EEG and video data from the closeness manipulation itself are reported elsewhere (Fornari et al. 2026). Because both papers are based on the same participant sample and share identical EEG and video acquisition and preprocessing procedures, parts of the methodological description in the Methods section overlap.

2.4. Manipulation

Each dyad was randomly assigned to one of three conditions: a No‐Interaction (NI) condition, in which partners silently thought about a set of superficial questions; a Small‐Talk (ST) condition, in which they asked each other the same superficial questions; and a Fast‐Friends (FF) condition, in which they asked each other a sequence of increasingly intimate questions specifically designed to increase interpersonal closeness between dyads of strangers (Fast‐Friends procedure; Aron et al. 1997; see Supporting Information D). According to previous findings (Aron et al. 1997; Sprecher 2021), dyads in the FF condition reliably report greater perceived closeness after the conversation than those in the ST condition, making this an ideal paradigm for investigating how variations in interpersonal closeness impact inter‐brain synchrony (IBS).

Both the superficial and intimate items were drawn from Aron's original Fast‐Friends procedure, with each set comprising 36 questions divided into three 12‐item lists. However, we used an abbreviated 24‐min protocol validated by Sprecher (2021), consisting of three 8‐min blocks, rather than Aron et al.'s original 45‐min version. This shorter version was used to reduce the total experimental duration, as the EEG gel can dry out during long recordings. In each block, participants received a question set; experimenters left the room, returning only to deliver the next set. In the ST and FF conditions, partners sat side‐by‐side, oriented at approximately 45° relative to the camera and positioned about 1.5 m away (Figure 1), alternated questions, and could skip items if they felt uncomfortable answering. Interactions were video‐recorded, and participants were instructed to keep the question sheets on their laps so that their upper bodies remained visible. In NI, a divider blocked eye contact, which is known to affect IBS (Koul et al. 2023). Audio was not recorded in this phase to safeguard participants' privacy and to avoid compromising the manipulation, as dyads in the FF condition were instructed to discuss increasingly intimate topics. Table 1 reports the composition of dyads in each condition by participants' language and gender, as dyads were matched on both characteristics.

TABLE 1.

Number of dyads by gender, language, and experimental condition.

Condition	Dutch		English		Total
	Female	Male	Female	Male
NI	16	4	16	5	41
ST	14	7	18	2	41
FF	16	4	16	5	41

Abbreviations: FF, Fast Friends condition; NI, No Interaction condition; ST, Small Talk condition.

2.5. Collaborative Task

Following the interpersonal closeness manipulation, dyads engaged in a computer‐mediated collaborative maze task (Healey and Mills 2006; Mills 2014, based on Garrod and Anderson 1987). Participants were seated on opposite sides of a table, facing each other and their laptops (Figure 2). The task comprised 10 mazes selected from Mills's 12‐maze TurnByTurn_Dyadic_Maze_Game set, presented via the Chattool Java application (https://dialoguetoolkit.github.io/chattool/). Although Chattool includes a text‐chat function, it was not used in the present study, as participants communicated orally.

The task consisted of two blocks of five mazes, separated by a two‐minute rest period. Participants could proceed to the next maze only after completing the preceding one; no interruptions occurred between mazes except for the scheduled break between the two blocks. All dyads completed all 10 mazes. Each maze was presented on a 7 × 7 grid. Both players viewed the same maze layout but had different starting positions, goal locations, gates, and switches, none of which were visible to the partner (Figure 3). Each player's position appeared only on their own screen. The path toward the goal was obstructed by gates that could be opened only when the partner stepped onto a node marked as a switch on the player's own screen. Thus, successful navigation required participants to verbally guide each other to relevant switches (e.g., Player 1 directing Player 2 to switches visible only to Player 1, and vice versa). The maze was completed only when both players reached their respective goal locations.

FIGURE 3.

Example of maze configurations for Players 1 (A) and 2 (B). The two mazes have an identical topology. The black circle indicates the player's current position, the cross indicates the goal position, solid bars represent gates, and shaded areas indicate switches that open the gates (adapted from Mills 2014).

Participants were instructed to solve each maze as quickly as possible, although no explicit time limit was imposed. Because fast completion was the stated goal of the task, collaborative performance was measured as total completion time across the task (Block 1 + Block 2), excluding the 2‐min break. This task was selected because it involves all the core features of collaborative interaction, including a shared goal, positive interdependence, and equal status and knowledge among partners (Dillenbourg 1999; Salomon and Globerson 1989). EEG, video, and audio were recorded during the collaborative task. Two experimenters were present in the room while each dyad solved the ten mazes, ensuring that the task rules were followed and that participants remained actively engaged throughout the task.

2.6. Perceived Collaboration Quality

Perceived collaboration quality was assessed with a single item administered at the end of the collaborative game (“How well do you think the collaboration went?”). Responses were provided on a sliding scale ranging from 0 (not well at all) to 100 (extremely well). Dyad‐level perceived collaboration quality was computed as the mean of the two partners' ratings.

2.7. EEG Recording

Brain activity was recorded with two 24‐channel (22 channels plus 2 mastoid channels) SMARTING mobile EEG systems (mBrainTrain, n.d.; 250 Hz sampling rate) mounted on gel‐based caps (EASYCAP RBE‐24‐STD). Electrodes were placed according to the international 10–20 system, with FCz serving as the reference (CMS) and Fpz as the ground (DRL). Before every session, all impedances were verified to be < 10 kΩ. Each wireless amplifier was clipped to the back of the cap, below the occipital electrodes, and streamed data via Bluetooth to a USB receiver fixed to the participant's chair; the receiver was connected by an extension cable to a dedicated Windows 10 laptop.

Although each participant's EEG was streamed to a separate laptop using the mbtStreamer software (mBrainTrain, n.d.), the actual recording was carried out on only one of the two laptops. This was possible because the two acquisition laptops were connected to the same local network via a router. Specifically, mbtStreamer uses Lab Streaming Layer (LSL) to insert precise timestamps into the data streams, allowing for subsequent correction for clock offsets between the two EEG recordings. Two recordings were taken for each dyad, one for each block of the collaborative task. The two EEG streams (from participants 1 and 2) of each block were saved in a single XDF file and imported into MATLAB (R2023a; The MathWorks Inc. 2023), where the MoBILAB toolbox (Ojeda et al. 2014) was used to re‐align them based on the LSL timestamps. The synchronized data were then exported in BrainVision Analyzer format for preprocessing.

2.8. EEG Preprocessing

EEG data from the 2 blocks of the collaborative task were preprocessed in BrainVision Analyzer v2.3 (Brain Products GmbH 2019). As suggested by the DEEP pipeline (Kayhan et al. 2022), temporal electrodes (F7, F8, T7, T8, P7, P8) were removed because speech‐related muscle artifacts rendered them unsuitable for analysis. Occipital electrodes (O1, O2) were also removed, as the amplifier positioned just below often caused poor scalp contact and frequent detachment. Preprocessing was therefore conducted on 14 channels: (Fp1, Fp2, Afz, F3, F4, Fz, C3, C4, Cz, P3, P4, Pz, CPz, POz).

Data were bandpass‐filtered between 0.5 and 40 Hz using a zero‐phase Butterworth filter (4th‐order high‐pass, 8th‐order low‐pass), and a notch filter (50 Hz) was applied to remove remaining line noise. Artifacts were first identified using automatic inspection (gradient 50 μV/ms; max‐min 100 μV; amplitude −100 to 100 μV; low activity 0.5 μV), then by visual inspection. ICA (Jung et al. 1998) was then used to remove blink and eye‐movement components. Noisy channels were excluded from ICA and later interpolated (spherical‐spline interpolation). In accordance with the DEEP guidelines, we aimed to interpolate no more than 10% of channels (i.e., a maximum of 2 out of 14). However, to maximize data retention, we exceeded this threshold in 15% of participants (3 channels in 30 participants; 4 channels in 7 participants). Data were then re‐referenced to the robust‐average reference (Kayhan et al. 2022), which excludes interpolated channels, and then inspected a second time using automatic data inspection. Finally, data were epoched into 4‐s segments and exported as DAT files. The 4‐s epoch length was chosen to reduce inflation of phase‐coupling estimates (Basti et al. 2022; Zimmermann et al. 2024).

2.9. EEG Analysis

Inter‐brain synchrony (IBS) was calculated in MATLAB (R2023a; The MathWorks Inc. 2023). The two EEG streams for each dyad and block (Block 1 and Block 2) were uploaded to MATLAB using the EEGLAB toolbox (Delorme and Makeig 2004). Using FieldTrip (Oostenveld et al. 2011), data were band‐pass filtered into delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), and beta (12–30 Hz) ranges. Instantaneous phase was then extracted from each band‐filtered signal via the Hilbert transform.

For every 4‐s epoch and every homologous electrode pair (e.g., Fp1‐Fp1, Fp2‐Fp2), IBS was calculated using the circular correlation coefficient (CCorr; Jammalamadaka and Sengupta 2001), the circular analogue of the Pearson Correlation Coefficient. We chose CCorr because it is robust against coincidental phase alignment and exhibits a lower Type‐1 error rate than many alternative IBS metrics (Burgess 2013). Of the two formulas for CCorr described by Jammalamadaka and Sengupta (2001), we used eq. 8.2.4, which is more accurate for data with uniform phase distributions, such as EEG signals, and arbitrary epoch‐wise mean phase directions (Zimmermann et al. 2024). To compute the adjusted CCorr for univariate data and assess phase uniformity, we used the MATLAB script provided by Zimmermann et al. (2024), available on GitHub (https://github.com/marizi/CCorrIBS).

Uniformity was tested using the Hodges‐Ajne omnibus test for nonuniformity (CircStat toolbox; function: circ_otest; Berens 2009). Data were uniform in most epochs for the theta (98.91%), alpha (100%), and beta (99.95%) frequency bands. However, delta‐band data showed predominantly non‐uniform phase distributions, with only 5.03% of epochs classified as uniform. This likely reflects the low number of complete oscillations in a 4‐s window at delta frequencies (e.g., a 0.5 Hz wave completes only 2 full cycles in 4 s). Given the high phase uniformity in theta, alpha, and beta bands, and under the assumption that the mean phase direction per epoch is arbitrary, we adopted the adjusted CCorr formula proposed by Jammalamadaka and Sengupta (2001; eq. 8.2.4). We applied the same formula to the delta band, as the data still exhibit arbitrary mean directions despite a non‐uniform phase distribution.

CCorr values were converted to absolute values (since both positive and negative values indicate synchrony) and Fisher Z‐transformed to approximate normality. Because EEG recording length varied across dyads, depending on how quickly they solved the ten mazes (five mazes in Block 1 and five in Block 2), the number of artifact‐free epochs also differed between dyads. To ensure a consistent number of epochs per dyad, we selected the minimum number of artifact‐free epochs observed across all dyads (considering both blocks) as the reference (n = 141 epochs). For each dyad, we then randomly sampled 141 artifact‐free epochs from the entire recording (Block 1 + Block 2). This procedure ensured that differences in circular correlation (CCorr) were not driven by an unequal number of epochs (Miljevic et al. 2022). The CCorr values were then averaged across the selected epochs, yielding one synchrony score per electrode pair. Finally, provided that CCorr values could vary between brain regions, similarly to Davidesco et al. (2023), we averaged CCorr values across electrode pairs within three regions of interest: Frontal (Fp1, Fp2, AFz, F3, F4, Fz), Central (C3, C4, Cz, CPz), and Posterior (P3, P4, Pz, POz).

As a sensitivity analysis, we addressed potential concerns that the high‐pass filter cutoff (0.5 Hz) might attenuate delta frequencies near the lower bound of the delta band (0.5–4 Hz) by repeating the analyses using a narrower delta range (1–4 Hz). Results reported in the main text are based on the 0.5–4 Hz band; findings from the sensitivity analysis are presented in Supporting Information B.

2.10. Video Analysis

Videos were recorded under constant lighting and preprocessed using Motion Energy Analysis (MEA; Ramseyer 2020; Supporting Information A3). Motor synchrony was then analyzed in R (v4.4.0) with the rMEA package (Kleinbub and Ramseyer 2021). Because gestures often occluded the face, motion‐energy values from the head and body ROIs were summed to create a single time series per participant. Each series was smoothed with a 0.4‐s moving average to suppress high‐frequency distortions (Tschacher et al. 2014) and then rescaled so that the following comparisons reflected relative, not absolute, movement amplitude. Then, extreme outliers (> ±10 SD) were removed. Following Asher et al. (2020), partners' time series were cross‐correlated within non‐overlapping 30‐s windows, with lags from −5 to +5 s in 0.04‐s steps (125 shifts per direction). Coefficients were Fisher‐Z‐transformed and converted to absolute values (Tschacher et al. 2014). Because video duration varied across dyads, depending on how quickly participants solved the ten mazes, we used the minimum number of 30‐s windows observed across dyads, considering both blocks (n = 18), as a reference. From each recording, we then randomly sampled this fixed number of 30‐s windows across the entire session (Block 1 + Block 2). This ensured that differences in motor synchrony were not driven by unequal recording durations.

Absolute values were used to ensure that any systematic temporal coupling (both positive and negative cross‐correlation) contributed positively to the synchrony measure (Ramseyer and Tschacher 2011). In addition, the ±5‐s lag window allowed the measure to capture not only simultaneous alignment but also delayed coordination characteristic of spontaneous motor mimicry, a hallmark of naturalistic interaction (Dumas and Fairhurst 2021). Together, the use of absolute values and time lags allows motor synchrony to be interpreted as a broader form of movement coordination and turn‐taking, rather than strictly simultaneous movement.

2.11. Statistical Analysis

2.11.1. Main Preregistered Analyses

All statistical analyses were conducted in R (version 4.4.0; R Core Team 2024). The main analyses were conducted on 119 dyads (40 FF, 39 ST, 40 NI), after excluding four dyads due to technical problems (Bluetooth disconnections or EEG cap malfunction) that resulted in EEG data loss and invalid task completion times.

2.11.1.1. Effect of Interpersonal Closeness on Collaborative Performance

The impact of interpersonal closeness on collaborative performance was analyzed with an ANOVA, with Condition (NI, ST, FF) as the main factor, Language (Dutch, English) as a control factor, and the dyad's task completion time as the outcome. In a subsequent exploratory step, a Condition × Language interaction term was added to test whether the effect of Condition was moderated by the language spoken within the dyad. Effect sizes were quantified as partial η ², computed from the linear model using the effectsize package (Ben‐Shachar et al. 2020). Post hoc tests used Tukey‐adjusted estimated marginal means.

2.11.1.2. Effect of Interpersonal Closeness on Inter‐Brain Synchrony During Collaboration

IBS values for each frequency band (delta, theta, alpha, beta), brain region (frontal, central, posterior), and condition (NI, ST, FF) were screened for extreme outliers (> 3 × IQR). No extreme outliers were found. ICCs indicated substantial dyad‐level clustering across repeated measurements from three brain regions for all frequency bands (delta = 0.50, theta = 0.32, alpha = 0.33, beta = 0.36); therefore, all models included a random intercept for dyad. For each frequency band, IBS was analyzed with a linear mixed‐effects model (lmer::lme4, Bates et al. 2025), with Condition (NI, ST, FF), Brain Region (Frontal, Central, Posterior), and their interaction as fixed effects, and the dyad's mean IBS during collaboration as the outcome. Language (Dutch, English) was included as a control predictor. p‐Values were obtained with lmerTest::anova (Kuznetsova et al. 2017), with Satterthwaite approximation for the degrees of freedom. Post hoc comparisons for any significant effects were conducted using estimated marginal means (Searle et al. 1980) (emmeans package; Lenth et al. 2025), with Satterthwaite approximation for degrees of freedom. Tukey adjustments were applied to post hoc comparisons of significant main effects. Significant Condition × Brain Region interactions were followed up by comparing conditions within each brain region. p‐Values for these follow‐up comparisons were adjusted using the false discovery rate (FDR) procedure. Effect sizes were not reported because there is no single standardized effect‐size measure for contrasts in mixed‐effects models, and available measures depend on choices about the appropriate variance component used for standardization (Rights and Sterba 2019).

2.11.1.3. Predictive Value of Inter‐Brain Synchrony for Collaborative Performance

To assess whether IBS during collaboration predicted collaborative performance above and beyond perceived collaboration quality, we conducted a two‐step hierarchical multiple regression across all dyads. In Step 1, task completion time was regressed on perceived collaboration quality, with language (Dutch vs. English; dummy‐coded) included as a control predictor. In Step 2, IBS values from the frontal, central, and posterior regions were added to the model. The two models were then compared (ΔR ², ΔF) to determine whether IBS significantly explained additional variance in collaborative performance beyond perceived collaboration quality. Effect sizes for each predictor were expressed as semi‐partial R ² (sr ²), reflecting the unique variance in the dependent variable explained by each predictor. This analysis was conducted separately for each frequency band (delta, theta, alpha, beta).

2.11.2. Secondary Preregistered Analyses

These analyses were conducted on all 123 dyads (41 FF, 41 ST, 41 NI).

2.11.2.1. Effect of Interpersonal Closeness on Motor Synchrony During Collaboration

The effect of interpersonal closeness on motor synchrony was analyzed using an ANOVA, with Condition (NI, ST, FF) as the independent factor, and the dyad's mean motor synchrony during the collaborative task as the outcome. Effect sizes were quantified as partial η ², computed from the linear model using the effectsize package (Ben‐Shachar et al. 2020). Post hoc tests used Tukey‐adjusted estimated marginal means.

2.11.2.2. Change in Affiliative Feelings (Interpersonal Closeness, Liking, and Perceived Similarity)

Dyad‐level outcomes (closeness, liking, and perceived similarity) were computed as the average of the two participants' ratings. Ratings were collected at three time points (pre‐manipulation, post‐manipulation, and post‐collaboration), yielding three observations per dyad. Random‐intercept‐only models showed substantial dyad‐level clustering across repeated measures of affiliative feelings for all outcomes (closeness = 0.30, liking = 0.32, similarity = 0.36); therefore, all models included a random intercept for dyad. Each outcome was analyzed with a linear mixed‐effects model (lmer::lme4, Bates et al. 2025) with Condition (NI, ST, FF), Time (T1 = pre‐manipulation, T2 = post‐manipulation, T3 = post‐collaboration), and their interaction as fixed effects. p‐Values were obtained with lmerTest::anova (Kuznetsova et al. 2017), with Satterthwaite approximation for the degrees of freedom. Post hoc comparisons for significant Condition × Time interactions were conducted using estimated marginal means (Searle et al. 1980) (emmeans package; Lenth et al. 2025), with Satterthwaite approximation for degrees of freedom. p‐Values were adjusted using the false discovery rate (FDR) procedure.

3. Results

3.1. Main Preregistered Analyses

3.1.1. Effect of Interpersonal Closeness on Collaborative Performance

No significant effect of Condition on collaborative performance was observed, F(2, 115) = 0.61, p = 0.547, η ² ₚ = 0.01. However, a significant effect of Language emerged, F(1, 115) = 24.13, p < 0.001, η ² ₚ = 0.17, indicating that Dutch native speaker dyads completed the task significantly faster (M = 1003 s, SD = 275 s) than English‐speaking dyads (M = 1285 s, SD = 344 s).

When the Condition × Language interaction term was included, no significant interaction was found, F(2, 113) = 0.71, p = 0.493, η ² ₚ = 0.01. The main effect of Condition remained non‐significant, F(2, 113) = 0.62, p = 0.541, η ² ₚ = 0.01, while the main effect of Language was maintained, F(1, 113) = 24.14, p < 0.001, η ² ₚ = 0.18. Mean, standard deviation, and range of maze task completion time across the three conditions are reported in Table 2.

TABLE 2.

Mean, standard deviation, and range of maze task completion time by condition, in minutes.

Condition	Mean time (min)	SD time (min)	Range	n
NI	19.57	5.68	11.47–34.53	40
ST	19.35	6.52	10.33–38.07	39
FF	18.32	4.82	9.67–30.07	40

Abbreviations: FF, Fast Friends condition; NI, No Interaction condition; ST, Small Talk condition.

3.1.2. Effect of Interpersonal Closeness on Inter‐Brain Synchrony During Collaboration

No significant main effect of Condition emerged in any of the examined frequency bands (all ps > 0.10). However, a significant main effect of Brain Region was observed in both the delta, F(2, 232) = 25.74, p < 0.001, and beta bands, F(2, 232) = 7.38, p < 0.001. Additionally, a significant Condition × Brain Region interaction was observed in the theta band, F(4, 232) = 2.99, p = 0.020. A complete overview of the results can be found in Table C1.

Tukey‐adjusted pairwise comparisons for the main effect of Brain Region revealed that in the delta band, IBS was significantly higher in both frontal and central regions compared to the posterior region (Frontal vs. Posterior: b = 0.005, SE = 0.001, t(232) = 5.81, p < 0.001; Central vs. Posterior: b = 0.005, SE = 0.001, t(232) = 6.55, p < 0.001), with no difference between frontal and central regions. In the beta band, the opposite pattern emerged: IBS was significantly higher in the posterior region compared to both frontal and central regions (Frontal vs. Posterior: b = −0.001, SE = 0.0003, t(232) = −3.50, p = 0.002; Central vs. Posterior: b = −0.001, SE = 0.0003, t(232) = −3.12, p = 0.006), with no significant difference between frontal and central regions. None of the post hoc comparisons for the Condition × Brain Region interaction in the theta band were significant after FDR correction across all interaction comparisons (all ps ≥ 0.10). Post hoc results for the delta, theta, and beta bands are reported in Table C2.

3.1.3. Predictive Value of Inter‐Brain Synchrony for Collaborative Performance

In Step1, perceived collaboration quality and language significantly explained variance in completion time, R ² = 0.26, F(2,116) = 20.19, p < 0.001. Better perceived collaboration quality was associated with shorter completion times, b = −8.89, SE = 2.43, t(116) = −3.66, sr ² = 0.09, p < 0.001 (Figure 4A), and English‐speaking dyads completed the task more slowly than Dutch‐speaking dyads, b = 133.42, SE = 27.23, t(116) = 4.90, sr ² = 0.15, p < 0.001. Adding IBS measures in Step 2 significantly improved model fit only in the delta frequency band, ΔR ² = 0.06, F‐change (3113) = 3.31, p = 0.023; no improvement was observed for theta, alpha, or beta bands. Multicollinearity was low across predictors (all VIFs ≤ 2.02). Within the delta band, higher central IBS was associated with longer completion times, b = 11,202.48, SE = 3697.55, t(113) = 3.03, sr ² = 0.06, p = 0.003 (Figure 4B). No significant IBS effects were found for the theta, alpha, or beta bands (all ps > 0.10). Complete regression results are reported in Table C3.

FIGURE 4.

Associations of perceived collaboration quality and central delta‐band IBS with completion time. Panel A shows that higher perceived collaboration quality is associated with shorter completion time. Panel B shows that higher central delta‐band IBS is associated with longer completion time. Points represent dyads; lines show least‐squares fit with 95% CI. IBS, inter‐brain synchrony.

3.2. Secondary Preregistered Analyses

3.2.1. Effect of Interpersonal Closeness on Motor Synchrony During Collaboration

No significant effect of Condition on motor synchrony was observed, F(2, 120) = 0.59, p = 0.555, η ² ₚ = 0.01.

3.2.2. Change in Affiliative Feelings (Interpersonal Closeness, Liking, and Perceived Similarity)

This analysis examined whether the closeness manipulation successfully induced differences in affiliative feelings between conditions and whether affiliative feelings further increased following collaboration. For each measure (interpersonal closeness, liking, and perceived similarity), ratings were collected at three time points: pre‐manipulation (T1), post‐manipulation (T2), and post‐collaboration (T3), yielding three repeated observations per dyad. Descriptive statistics are reported in Table C5 and post hoc comparisons in Table C6.

3.2.2.1. Interpersonal Closeness

For interpersonal closeness, we observed a significant main effect of Condition, F(2, 120) = 17.85, p < 0.001, and Time, F(2, 240) = 309.68, p < 0.001, as well as a significant Condition × Time interaction, F(4, 240) = 29.32, p < 0.001 (Figure 5).

FIGURE 5.

Change in reported interpersonal closeness over time across the three experimental conditions. FF, Fast Friends; NI, No Interaction; ST, Small Talk. Interpersonal closeness was measured before the manipulation (Time Point 1), after the manipulation (Time Point 2), and after the collaborative task (Time Point 3). Error bars represent the standard error of the mean (±SE). Asterisks indicate significant pairwise comparisons: *p < 0.05, ***p < 0.001.

3.2.2.1.1. Manipulation Effect (T2 vs. T1)

Interpersonal closeness increased from T1 to T2 in both interactive conditions, ST (b = 1.29, SE = 0.11, t(240) = 12.19, p < 0.001) and FF (b = 1.72, SE = 0.11, t(240) = 16.22, p < 0.001), but not in the NI condition (b = 0.18, SE = 0.11, t(240) = 1.73, p = 0.103). Baseline closeness did not differ across conditions. At T2, closeness was higher in both interactive conditions than in NI (NI vs. ST: b = −1.24, SE = 0.19, t(189) = −6.71, p < 0.001; NI vs. FF: b = −1.68, SE = 0.19, t(189) = −9.07, p < 0.001), and was higher in FF than in ST, ST vs. FF: b = −0.44, SE = 0.19, t(189) = −2.37, p = 0.028.

3.2.2.1.2. Collaboration Effect (T3 vs. T2)

From T2 to T3, closeness further increased in NI (b = 0.85, SE = 0.11, t(240) = 8.05, p < 0.001) and FF (b = 0.23, SE = 0.11, t(240) = 2.19, p = 0.045), but not in ST (b = 0.15, SE = 0.11, t(240) = 1.38, p = 0.169). At T3, the post‐manipulation pattern remained: closeness was higher in both interactive conditions than in NI (NI vs. ST: b = −0.54, SE = 0.19, t(189) = −2.89, p = 0.009; NI vs. FF: b = −1.06, SE = 0.19, t(189) = −5.72, p < 0.001) and remained higher in FF than ST, ST vs. FF: b = −0.52, SE = 0.19, t(189) = −2.83, p = 0.009.

In addition to interpersonal closeness, which was the main focus of this study, we examined two further affiliative measures (liking and perceived similarity), consistent with Sprecher (2021).

3.2.2.2. Liking

For liking, we observed a significant main effect of Condition, F(2, 120) = 6.62, p = 0.002, and Time, F(2, 240) = 183.51, p < 0.001, as well as a significant Condition × Time interaction, F(4, 240) = 28.49, p < 0.001 (Figure 6).

FIGURE 6.

Change in reported liking over time across the three experimental conditions. FF, Fast Friends; NI, No Interaction; ST, Small Talk. Liking was measured before the manipulation (Time Point 1), after the manipulation (Time Point 2), and after the collaborative task (Time Point 3). Error bars represent the standard error of the mean (±SE). Asterisks indicate significant pairwise comparisons: ***p < 0.001.

3.2.2.2.1. Manipulation Effect (T2 vs. T1)

Liking increased from T1 to T2 in ST (b = 1.17, SE = 0.11, t(240) = 11.12, p < 0.001) and FF (b = 1.38, SE = 0.11, t(240) = 13.09, p < 0.001), but not in NI (b = −0.02, SE = 0.11, t(240) = −0.23, p = 0.980). Baseline liking did not differ across conditions. At T2, liking was higher in both interactive conditions than in NI (NI vs. ST: b = −1.02, SE = 0.18, t(195) = −5.77, p < 0.001; NI vs. FF: b = −1.39, SE = 0.18, t(195) = −7.83, p < 0.001), whereas ST and FF did not differ significantly, ST vs. FF: b = −0.37, SE = 0.18, t(195) = −2.06, p = 0.122.

3.2.2.2.2. Collaboration Effect (T3 vs. T2)

Liking increased from T2 to T3 in NI (b = 0.95, SE = 0.11, t(240) = 9.04, p < 0.001) but did not change in ST (b = 0.00, SE = 0.11, t(240) = 0.00, p = 1.00) or FF (b = −0.12, SE = 0.11, t(240) = −1.16, p = 0.372). At T3, differences in liking between interactive and non‐interactive conditions were no longer significant (NI vs. ST: b = −0.07, SE = 0.18, t(195) = −0.41, p = 0.766; NI vs. FF: b = −0.32, SE = 0.18, t(195) = −1.79, p = 0.171; ST vs. FF: b = −0.24, SE = 0.18, t(195) = −1.37, p = 0.308).

3.2.2.3. Perceived Similarity

We observed a significant main effect of Condition, F(2, 120) = 11.80, p < 0.001, Time, F(2, 240) = 193.32, p < 0.001, and a Condition × Time interaction, F(4, 240) = 18.17, p < 0.001 (Figure 7).

FIGURE 7.

Change in reported perceived similarity over time across the three experimental conditions. FF, Fast Friends; NI, No Interaction; ST, Small Talk. Perceived Similarity was measured before the manipulation (Time Point 1), after the manipulation (Time Point 2), and after the collaborative task (Time Point 3). Error bars represent the standard error of the mean (±SE). Asterisks indicate significant pairwise comparisons: *p < 0.05, **p < 0.01, ***p < 0.001.

3.2.2.3.1. Manipulation Effect (T2 vs. T1)

Perceived similarity increased from T1 to T2 in ST (b = 1.17, SE = 0.13, t(240) = 9.27, p < 0.001) and FF (b = 1.59, SE = 0.13, t(240) = 12.55, p < 0.001), but not in NI (b = 0.12, SE = 0.13, t(240) = 0.97, p = 0.402). Baseline perceived similarity did not differ across conditions. At T2, similarity was higher in both interactive conditions than in NI (NI vs. ST: b = −0.84, SE = 0.21, t(196) = −3.96, p < 0.001; NI vs. FF: b = −1.67, SE = 0.21, t(196) = −7.86, p < 0.001), and was higher in FF than ST, ST vs. FF: b = −0.83, SE = 0.21, t(196) = −3.90, p < 0.001.

3.2.2.3.2. Collaboration Effect (T3 vs. T2)

From T2 to T3, similarity increased in NI, b = 0.96, SE = 0.13, t(240) = 7.63, p < 0.001, and ST, b = 0.31, SE = 0.13, t(240) = 2.41, p = 0.025, but not in FF, b = 0.06, SE = 0.13, t(240) = 0.48, p = 0.630. At T3, similarity remained higher in FF than in NI and ST (NI vs. FF: b = −0.77, SE = 0.21, t(196) = −3.61, p = 0.001; ST vs. FF: b = −0.59, SE = 0.21, t(196) = −2.75, p = 0.012), whereas the NI‐ST difference was no longer significant, NI vs. ST: b = −0.18, SE = 0.21, t(196) = −0.86, p = 0.391.

3.3. Additional Non‐Preregistered Analyses

The following analyses were conducted on 119 dyads (40 FF, 39 ST, 40 NI), after excluding four dyads due to technical problems (Bluetooth disconnections or EEG cap malfunction) that resulted in EEG data loss and invalid task completion times.

3.3.1. Sensitivity Analysis Excluding More Acquainted Dyads

The three main preregistered analyses were re‐run after excluding the four dyads with acquaintance scores above 20 (on a scale from 0 to 100), to assess whether these dyads influenced the results. The pattern of results remained unchanged. Specifically, there was no effect of condition on collaborative performance, operationalized as completion time, or on IBS in any frequency band (all ps > 0.10). Moreover, the positive association between central delta IBS and collaborative performance remained significant, p = 0.003. These results suggest that the inclusion of the four dyads with higher acquaintance scores did not drive the main findings.

3.3.2. Correlation Between Central Delta IBS and Motor Synchrony

To examine whether motor synchrony was associated with central delta IBS, we tested the dyad‐level relationship between these variables using a Pearson correlation. No association was observed, r(117) = −0.03, p = 0.759.

3.3.3. Correlation Between Motor Synchrony and Performance

The association between motor synchrony and task completion time was examined using a Pearson correlation. Motor synchrony was positively correlated with completion time, indicating that higher synchrony was associated with longer completion times, r(117) = 0.22, p = 0.017 (Figure 8).

FIGURE 8.

Association between motor synchrony and completion time. Higher motor synchrony is associated with longer completion time. Points represent dyads; lines show least‐squares fit with 95% CI.

3.3.4. Predictive Value of Inter‐Brain Synchrony for Performance Measured as Number of Moves

As an additional analysis, we assessed whether IBS during collaboration predicted the total number of moves made by each dyad to complete the collaborative maze task. To test this, we ran four multiple regression models, one for each frequency band (delta, theta, alpha, beta). The total number of moves was regressed on perceived collaboration quality and IBS values from frontal, central, and posterior regions. Language (Dutch vs. English; dummy‐coded) was included as a control predictor. Effect sizes for each predictor were expressed as semi‐partial R ² (sr ²), reflecting the unique variance in the dependent variable explained by each predictor. Within the theta band, higher frontal IBS was associated with a higher number of moves, b = 6818.71, SE = 2302.89, t(113) = 2.96, sr ² = 0.07, p = 0.004. No significant IBS effects were found for the delta, alpha, or beta bands (all ps > 0.10).

3.3.5. Predictive Value of Collaboration Time for Post‐Collaboration Affiliative Feelings

This analysis examined whether longer maze completion times, and therefore longer collaboration, were associated with higher post‐collaboration affiliative feelings. We fitted separate multiple regression models predicting post‐collaboration interpersonal closeness, liking, and perceived similarity. Each model included completion time, Condition (NI, ST, FF), and their interaction as predictors, with the corresponding pre‐collaboration affiliative measure included as a control predictor.

After accounting for pre‐collaboration scores, completion time did not significantly predict post‐collaboration interpersonal closeness, liking, or perceived similarity (all ps ≥ 0.10). The Completion Time × Condition interactions were also nonsignificant (all ps ≥ 0.10), providing no evidence that these associations varied by condition. Thus, dyads who collaborated for longer did not report significantly stronger post‐collaboration affiliative feelings.

3.3.6. IBS Changes Between Block 1 and Block 2 of the Collaborative Game

In this additional analysis, we examined whether IBS changed between Block 1 and Block 2 of the collaborative game. To do so, we sampled the same number of artifact‐free epochs from Block 1 and Block 2 for each dyad. The number of sampled epochs was determined by the minimum number of available artifact‐free epochs across blocks and dyads, which was 54 epochs in Block 2. Thus, 54 artifact‐free epochs were sampled from each block for each dyad. For each frequency band, IBS was analyzed using a linear mixed‐effects model with Condition (NI, ST, FF), Block (Block 1, Block 2), and their interaction as fixed effects. Mean IBS per dyad and block was used as the outcome variable, and all models included a random intercept for dyad. p‐Values were obtained with lmerTest::anova (Kuznetsova et al. 2017), with Satterthwaite approximation for the degrees of freedom. Post hoc comparisons for any significant effects were conducted using estimated marginal means (Searle et al. 1980) (emmeans package; Lenth et al. 2025), with Satterthwaite approximation for degrees of freedom. Tukey adjustments were applied to post hoc comparisons of significant main effects.

No main effect of Block and no Block × Condition interaction were found for the alpha, beta, or delta frequency bands. In the theta band, however, there was a significant main effect of Block, F(1, 119) = 4.08, p = 0.046. Estimated marginal means indicated that theta IBS was slightly higher in Block 2 than in Block 1, b = −0.001, SE = 0.00066, t(119) = −2.02, p = 0.046.

4. Discussion

4.1. Effects of Interpersonal Closeness on Collaboration and Inter‐Brain Synchrony

The main aim of the current experimental study was to investigate the causal effect of interpersonal closeness on inter‐brain synchrony (IBS) and performance during collaboration. Previous research has shown that individuals in closer relationships (e.g., romantic partners vs. strangers) tend to collaborate more efficiently and exhibit greater IBS during joint tasks (Djalovski et al. 2021; Pan et al. 2017). More recently, Zhang et al. (2024) reported higher IBS in friends compared to stranger dyads, although no corresponding differences in collaborative performance were observed. However, relationship types differ not only in interpersonal closeness but also in factors such as shared collaborative history and familiarity. To disentangle the specific effect of closeness from these potential confounds and to test for a causal relationship, we experimentally manipulated interpersonal closeness and examined its direct impact on IBS and collaboration.

The manipulation was effective: after the manipulation phase, dyads in the closeness‐generating conversation condition (FF) reported higher interpersonal closeness than those in the Small‐Talk condition (ST), and both interactive conditions (FF and ST) reported higher closeness than NI. Moreover, from pre‐ to post‐collaboration, closeness further increased in FF and NI, but not in ST. This increase in closeness following collaboration is consistent with prior evidence that collaboration can promote affiliation (Depping and Mandryk 2017; Johnson et al. 2014; Patané et al. 2020; Plötner et al. 2015; Yu et al. 2022). Notably, the increase in closeness was largest in dyads that had not previously interacted (NI), resulting in a reduced post‐collaboration difference between interacting (FF, ST) and non‐interacting (NI) dyads (see Figure 5).

In contrast with our hypotheses, the manipulation of interpersonal closeness did not affect collaborative performance or IBS during collaboration. This null effect suggests that the differences in IBS and collaborative performance observed between relationship types in previous studies (Brennan and Enns 2015; Djalovski et al. 2021; Jehn and Shah 1997; Pan et al. 2017) may be driven by factors other than a transient feeling of closeness. These may include familiarity between interaction partners and shared interaction history, as well as interaction‐related processes such as greater accuracy in predicting a partner's speech or actions, differences in engagement or motivation, and motor synchrony. Alternatively, although the manipulation effectively increased perceived closeness, the magnitude of the differences between conditions may have been insufficient to impact behavior or IBS. Moreover, it is important to consider that collaboration itself enhanced affiliation within dyads, especially in the NI condition, reducing differences between conditions and further decreasing the likelihood of detecting effects on IBS and collaborative performance.

4.2. Central Delta IBS During Collaboration Predicts Performance

Central delta IBS negatively predicted collaborative performance: dyads who took longer to complete the collaborative task also showed higher IBS (see Figure 4B). Importantly, IBS explained unique variance in performance beyond self‐reported perceptions of collaboration quality.

The counterintuitive pattern observed suggests that higher IBS may not always reflect more effective collaboration. Although most studies on collaborative interactions have reported a positive association between IBS and collaborative performance (Cui et al. 2012; Djalovski et al. 2021; Pan et al. 2017; Reinero et al. 2021; Wikström et al. 2022; Zhang et al. 2024), a few studies have found the opposite in contexts involving behavioral misalignment or increased coordination effort. For instance, Lender et al. (2023) found that theta‐ and delta‐band IBS between piano duet partners increased when their behavioral synchrony was disrupted. Similarly, Kurihara et al. (2022) reported that during anti‐phase finger tapping at high tempos, theta IBS in temporal and central regions was negatively associated with behavioral precision. These findings suggest that delta and theta IBS may reflect top‐down control processes required to maintain coordination, such as prediction and task monitoring, rather than simply reflecting coordination success. This interpretation is consistent with prior work linking theta and delta oscillations in individuals to selective attention to visual and auditory stimuli (Knyazev 2012), and more specifically to divided attention during audiovisual integration (Yang et al. 2024). In our task, participants were required to divide their attention between a visual stimulus (maze navigation) and an auditory stimulus (instructions provided by their partner) in order to achieve a shared goal. It is therefore possible that higher delta IBS in our paradigm reflects the increased cognitive effort required to establish and maintain a shared framework for coordinating navigation within the maze.

Another, not mutually exclusive, interpretation is that the negative association between delta IBS and performance might be driven by speech tracking. Prior studies have shown that listeners' delta‐band activity entrains to acoustic features of speech (Guo et al. 2025; Guo and Hsieh 2025; Mai and Wang 2023; Van Hirtum et al. 2023). In particular, delta‐band oscillations show enhanced entrainment to the envelope of meaningful speech relative to control conditions, suggesting a sensitivity of delta activity to higher‐order linguistic processing (Molinaro and Lizarazu 2018). Because speakers also hear their own voice, shared auditory entrainment can produce zero‐lag coupling between speaker and listener (Pérez et al. 2017).

Although speculative, dyads who struggled to complete the maze may have experienced greater communication difficulties and engaged in more verbal exchanges to coordinate their actions, leading to higher speech‐driven delta IBS. In contrast, dyads who established a clearer shared linguistic framework may have required fewer clarifications, completed the task more efficiently, and exhibited lower delta IBS. The fact that the association between performance and delta IBS was localized over central regions further reinforces the speech‐tracking interpretation, as motor and premotor cortices have been implicated in sentence‐level speech tracking and comprehension (Keitel et al. 2018; Morillon et al. 2019).

When the lower delta range (0.5–1 Hz) was excluded (see Supporting Information B2), the association between delta IBS and performance was no longer observed. One possible explanation is that, while aiming to avoid potential attenuation near the high‐pass filter cutoff, we removed a portion of the delta band specifically linked to speech tracking at the phrasal level (Bourguignon et al. 2020; Pérez‐Navarro et al. 2024).

In the present task, we considered completion time to be the most valid measure of collaborative performance because participants were explicitly instructed to solve each maze as quickly as possible. As an alternative index of task efficiency, we also considered the number of moves dyads made to complete the task. Although participants were not explicitly instructed to minimize the number of moves, a higher number of moves may reflect lower efficiency in solving the mazes. This additional analysis showed that frontal theta IBS positively predicted the number of moves. This finding is consistent with the direction of our primary result, namely, that higher central delta IBS was associated with longer completion time. Thus, the association between higher IBS and poorer collaborative performance was not limited to completion time, but also extended to an alternative measure of task efficiency, providing further support for this finding.

4.3. Additional Correlational Findings

Similar to IBS, motor synchrony was negatively associated with task performance (see Figure 8). A comparable interpretation may apply: dyads who struggled to complete the maze may have relied more on gestures, in combination with speech, to convey spatial information relative to their position in the maze or the location of switches and goals. A higher rate of gestural turn‐taking between partners may, in turn, have resulted in greater motor synchrony.

Interestingly, although both delta‐band IBS and motor synchrony were higher in worse‐performing dyads, the two measures were not correlated. This result is in line with Gordon et al.'s (2025) view of interpersonal synchrony as a multimodal phenomenon. In other words, motor and neural synchrony might capture different aspects of interpersonal synchrony.

4.4. Collaboration Increases Affiliative Feelings (Interpersonal Closeness, Liking, and Perceived Similarity)

As a secondary, non‐preregistered analysis, we examined whether collaboration increased affiliative feelings. Prior research has shown that dyadic collaboration and interdependence can enhance trust and affiliation, compared to competition or working alone (Depping and Mandryk 2017; Johnson et al. 2014; Johnson and Johnson 2005). However, relatively few studies have directly examined changes in affiliative measures in adults before and after collaboration with a stranger (Henderson et al. 2024; Patané et al. 2020). Patané et al. reported that interpersonal closeness increased from before to after collaboration, whereas Henderson et al. found such increases only in one of the two types of collaborative activities they tested. Building on this work, the present study provides evidence that collaborative interaction can increase affiliative feelings between previously unacquainted young adults.

Importantly, the pattern of affiliation change differed by condition. In the NI condition, where participants had no prior interaction, all three affiliative measures increased significantly. In the FF condition, where dyads had already engaged in a deeper conversation before collaborating, only interpersonal closeness showed a further significant increase during collaboration. In the ST condition, where dyads had engaged in Small Talk prior to collaboration, only perceived similarity increased (see Figures 5, 6, 7).

These increases, however, did not appear to depend on collaboration duration: after accounting for pre‐collaboration affiliation, duration was not significantly associated with any post‐collaboration affiliative measure, and these associations did not differ significantly by condition. One possible explanation is that a longer completion time not only provides more opportunity for affiliation to develop but may also indicate greater difficulty with coordination or collaboration. These difficulties could counteract any potential positive effect of longer interaction time on affiliation. However, this interpretation remains speculative.

4.5. Strengths, Limitations, Future Studies

The main strength of our design lies in its causal approach, which is quite rare in hyperscanning research (Schilbach and Redcay 2025), and allowed us to test the effect of interpersonal closeness on IBS and collaboration, disentangling it from the influence of familiarity. In addition, our sample size (N = 246, 123 dyads, 41 per condition) is larger than that of many hyperscanning studies, which frequently employ small samples because of the technical and logistical complexity of the experimental setup (Réveillé et al. 2024; Zhao et al. 2024).

Nevertheless, this study also presents several limitations. First, we cannot be certain that the difference in interpersonal closeness induced by our manipulation was strong enough to elicit measurable effects on IBS and collaboration. In fact, designing manipulations of interpersonal closeness that are sufficiently powerful to influence behavioral and neural outcomes without inducing familiarity remains extremely challenging. For example, including multiple sessions could strengthen closeness, but would also introduce potential confounds related to repeated interactions. Balancing these factors will be fundamental for isolating the causal effects of closeness on collaborative processes. A further challenge in studying the effects of experimentally induced closeness on collaboration among strangers is that collaboration itself promotes closeness, making it difficult to disentangle the effects of the manipulation from those of the collaborative experience.

A second methodological limitation of our study concerns the exclusion of all temporal electrodes due to muscle artifacts. This is not ideal for a study targeting IBS during verbal social interaction, given the role of temporal and inferior frontal cortices in speech processing and mentalizing. However, this exclusion was a necessary compromise to measure brain activity during an ecologically valid face‐to‐face interaction. This limitation could be partly addressed by using alternative mobile neuroimaging methods such as functional near‐infrared spectroscopy (fNIRS), which is less sensitive to muscle artifacts and offers higher spatial resolution. However, fNIRS also has important limitations, such as lower temporal resolution than EEG and reliance on indirect haemodynamic markers of neural activity. Given the complementary strengths and weaknesses of EEG and fNIRS, multimodal approaches combining both techniques may offer important advantages for future research (Li et al. 2022).

With respect to our findings, further research is needed to clarify the mechanisms underlying the observed negative association between delta‐band IBS and performance. As discussed above, this relationship could reflect increased speech production (and, consequently, stronger speech entrainment) in dyads that struggled more with the task. Alternatively, it may reflect stronger engagement of top‐down cognitive processes when coordination demands are high. Notably, the disappearance of the effect after excluding the 0.5–1 Hz interval is more consistent with a speech‐tracking interpretation.

To disentangle these possibilities (i.e., speech entrainment versus increased top‐down cognitive control), future studies should directly examine the relationship between IBS, speech‐envelope coupling, and higher‐order cognitive processes. In particular, experimental manipulations that disrupt verbal communication could be used to test whether IBS reflects cognitive effort for mutual understanding and behavioral coordination. This approach is conceptually similar to the one adopted by Lender and colleagues in their study of musical coordination, in which a confederate deliberately disrupted temporal alignment, leading to increased delta‐ and theta‐band IBS between players (Lender et al. 2023).

At the same time, we cannot exclude the possibility that the observed association between IBS and performance partly reflects differences in interaction duration. Dyads who took longer to complete the task necessarily spent more time interacting, and higher IBS may therefore partly reflect longer exposure to the collaborative interaction rather than poorer collaborative performance per se. This interpretation is consistent with evidence suggesting that collaborative interaction itself can contribute to increases in IBS (Khalil et al. 2022; Yun et al. 2012). However, our design does not allow us to directly disentangle performance from interaction duration, which represents an important limitation of the study. To partly address this limitation, we conducted an additional analysis using the number of moves as an alternative measure of task efficiency. Although minimizing the number of moves was not an explicit aim of the task, a higher number of moves may still indicate less efficient collaboration. This additional analysis revealed a positive association between frontal theta IBS and the number of moves. This finding strengthens our conclusions by showing that the association between higher IBS and poorer performance is not limited to completion time as the performance measure.

More generally, and independently of this limitation, the correlational nature of our findings means that the observed association between central delta‐band IBS and completion time cannot be interpreted as evidence for a specific causal effect of stronger speech tracking or greater cognitive effort in dyads who took longer. Rather, any explanation of the processes underlying this association remains speculative.

Beyond the present findings, a broader theoretical issue in the hyperscanning literature concerns the wide heterogeneity of tasks labeled as “collaborative”. This heterogeneity partly reflects the absence of a widely adopted, shared definition of collaboration, as well as the fact that collaboration can take multiple distinct forms. Paradigms labeled as collaborative range from highly constrained paradigms such as coordinated button‐pressing (Kruppa et al. 2021; Pan et al. 2017), which primarily capture motor coordination, to more interactive activities such as joint drawing (Djalovski et al. 2021), collaborative car racing (Wikström et al. 2022), tangram puzzle solving (Zhang et al. 2024), decision‐making (Zhang et al. 2021), creative design (Li et al. 2021), and problem‐solving (Reinero et al. 2021).

A related issue concerns how collaborative performance is operationalized. Most studies quantify performance using completion time (Wikström et al. 2022) or task scores (Pan et al. 2017; Reinero et al. 2021; Zhang et al. 2024). However, task scores often reflect different underlying behavioral components across paradigms, limiting comparability across studies. To capture collaboration quality more comprehensively, rather than just looking at performance, some hyperscanning studies use observer‐rated interaction dynamics (e.g., reciprocity, mutual adaptation, flow of ideas, or relaxed mood; Djalovski et al. 2021; Li et al. 2021), but these approaches also vary widely across tasks and rating criteria, again limiting comparability.

Our study focused on task performance, operationalized as completion time, because the task goal was to solve the mazes as quickly as possible, making speed an objective index of success. We found no evidence that interpersonal closeness affected completion time. However, completion time captures only how fast dyads are, and may be insensitive to more qualitative aspects of collaboration. Therefore, we cannot rule out that the closeness manipulation influenced collaboration‐relevant processes, such as coordination smoothness or the overall interaction atmosphere (e.g., warmth, engagement), which were not assessed and fall outside the scope of the current study.

Taken together, both diversity in task design and differences in the evaluation of collaboration likely contribute to the heterogeneity of IBS findings across studies (Réveillé et al. 2024). In line with this, recent evidence suggests that different forms of collaborative interaction may elicit different levels of IBS, with IBS being significantly higher during brainstorming than during problem‐solving (Azhari et al. 2025). Accordingly, rather than treating IBS as a broad marker of collaboration, it may be more informative to relate IBS to specific forms of collaborative activity and clearly defined performance or quality metrics.

Building on this insight, future studies should aim to clarify the multiple factors associated with IBS, from low‐level neural entrainment to higher‐order cognitive, social, and affective processes, to advance theoretical understanding of this still poorly understood phenomenon. Crucially, progress in this field will depend on moving beyond purely correlational designs to establish a deeper understanding of the causal relationships between neural and behavioral coordination.

5. Conclusion

Contrary to our predictions, experimentally inducing interpersonal closeness did not improve collaborative performance or increase IBS during collaboration. This suggests that the better collaboration and higher IBS observed in dyads with closer pre‐existing social bonds are not solely attributable to transient feelings of closeness. Instead, they may reflect other social factors, such as familiarity, as well as interaction‐specific processes such as engagement, mutual prediction, and motor synchrony. Future research should therefore aim to test the robustness of these null findings and use causal designs to disentangle these different possibilities.

Nevertheless, we observed a negative association between central delta IBS and collaborative performance, with IBS explaining additional variance beyond dyads' self‐reported performance. While this finding contradicts previous research, it may indicate that the IBS observed here reflects increased cognitive demands during coordination, such as divided attention and/or enhanced speech tracking, rather than coordination efficiency. The mechanisms underlying higher delta IBS in dyads with poorer performance remain to be elucidated.

Together, these findings contribute to the hyperscanning literature by promoting a causal approach, highlighting the need to disentangle affective‐emotional factors from familiarity, and suggesting that IBS is not always a marker of more efficient coordination, but may instead play a functional role in the emergence of behavioral coordination.

Author Contributions

Laura Fornari: Conceptualization, investigation, data curation, formal analysis, methodology, project administration, visualization, writing – original draft, writing – review and editing. Tieme W. P. Janssen: Conceptualization, funding acquisition, methodology, project administration, resources, software, supervision, writing – review and editing. Ido Davidesco: Conceptualization, methodology, software, supervision, writing – review and editing. Lydia Krabbendam: Conceptualization, funding acquisition, supervision, writing – review and editing. Nienke M. van Atteveldt: Conceptualization, funding acquisition, methodology, project administration, supervision, writing – review and editing.

Funding

This research was funded by the Ammodo Science Award 2020 for Social Sciences.

Disclosure

Open Science Statement: This study was preregistered on OSF: https://osf.io/vyt6n/overview?view_only=24c144d2f5ef49d9863aea6b7aed5fcb.

Ethics Statement

The study was approved by the ethics committee of the Faculty of Behavioral and Movement Sciences, Vrije Universiteit Amsterdam (VCWE‐2023‐061R1), and complied with the General Data Protection Regulation (GDPR).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information A: Experimental Procedures.

Supporting Information B: Analyses of the Delta Band in the 1‐4 Hz range.

Table C1: Main results of the mixed models for the effect of interpersonal closeness on IBS.

Table C2: Post hoc comparisons in the delta, beta, and theta frequency bands.

Table C3: Predictive value of IBS on collaborative performance.

Table C4: Predictive value of delta IBS (1–4 Hz) on collaborative performance.

Table C5: Average reported interpersonal closeness, liking, and perceived similarity pre‐manipulation (T1), post‐manipulation (T2), and post‐collaboration (T3).

Table C6: Post hoc comparisons for interpersonal closeness, liking, and perceived similarity.

Supporting Information D: Questions of the Fast Friends Procedure.

Data Availability Statement

Anonymized questionnaires, pre‐processed EEG, and motion energy data are available on Open Science Framework (OSF): https://osf.io/mbg7p/overview?view_only=c1ed26317e7046ff9fb3a5a2bff248ea. Raw EEG data and video recordings are available upon request due to participant confidentiality.