Empathy aligns brains in synchrony

Summary

Empathy is a core human capacity that underpins social life. Utilizing hyperscanning electroencephalography (EEG), we tested how empathy to others’ distress synchronizes brains without social cues. Mothers and adolescents (study 1, n = 100) underwent empathy-to-distress paradigm in separate rooms. Event-related interbrain connectivity was computed in four 500 ms time-windows following exposure to empathy-to-distress and control stimuli. Interbrain synchrony of fronto-temporal regions emerged throughout (0–2,000 ms) in alpha and beta bands following empathy-to-distress relative to control. Beta interbrain synchrony increased at 1,000–1,500 post-stimuli, indicating neural coupling of higher-order cognitive empathy. Oxytocin and behavioral synchrony correlated with enhanced interbrain synchrony. Study 2 replicated the paradigm with unacquainted adults (n = 44) and found interbrain beta synchrony for empathy stimuli at 1,000–1,500 ms post-stimulus. Exposure to others’ distress aligns brains in synchrony. Such alignment may have supported the consolidation of humans into social groups, increased affiliation and trust, and improved joint action to threats, enhancing survival and thriving.

Graphical abstract

Highlights

• •Interbrain synchrony in frontotemporal regions increased during empathy processing without social cues
• •Mother-child pairs had higher interbrain synchrony to empathy vs control images than strangers
• •Oxytocin and behavioral synchrony linked with greater interbrain synchrony to empathy cues
• •Beta synchrony at 1000–1500ms indicates higher-order processes in two-brain mechanisms of empathy

Cognitive neuroscience; Psychology

Introduction

The unprecedented dominance of Homo sapiens on Earth has been attributed to our species’ distinctive cognitive and social capabilities, which facilitate the execution of complex tasks through collaborative efforts.¹ The evolution of human societies capable of functioning as cohesive units requires not only common objectives but also the development of sophisticated mechanisms for real-time mutual understanding.² Central to these mechanisms is empathy, the capacity to comprehend and share others’ emotional states, which is fundamental to human social life.³ Empathy has been the subject of extensive research and provides a substrate for a wide array of prosocial behaviors. Empathy facilitates cooperation, sharing, and motivates altruistic actions, playing a key role in the development of complex human societies.⁴

Cognitive and affective empathy as complementary processes

The human ability to empathize is a complex and multilayered construct characterized by two distinct yet interrelated components: affective empathy and cognitive empathy. Affective empathy involves the automatic, often unconscious sharing of another person’s emotional state and resonance with the other’s experiences.⁵ While affective sharing may occur automatically, empathic responses have been found to be highly flexible and can be significantly modulated by contextual factors such as the interpersonal relationship between empathizer and target, or the adopted perspective of the observer.³ In contrast, cognitive empathy or perspective-taking relies on higher-order mentalizing processes that enable individuals to infer and evaluate the beliefs, goals, and intentions of others, without necessarily sharing their emotions.⁶

While these two types of empathy processes engage different neural systems, they often dynamically interact in real-world social interactions. This makes categorical distinction between the two processes difficult, with an ongoing debate whether affective and cognitive empathy represent points on a spectrum rather than discrete processes.⁷ However, while these two components are interrelated and operate in tandem to achieve behavioral goals, a substantial body of evidence suggests that affective and cognitive empathy implicate separate yet interconnected systems both behaviorally and neurally.⁸ Neuroimaging studies identified distinct neural networks associated with each component^9,10,11,12 and lesion studies provided evidence for their dissociation. Patients with ventromedial prefrontal lesions exhibited impaired cognitive empathy and retained intact affective empathy, whereas those with inferior frontal gyrus lesions displayed the inverse pattern,¹⁰ supporting the notion of their distinct function.

Further support for the distinction comes from studies of high-risk populations. Adults with autism spectrum disorders exhibit deficits in cognitive but not affective empathy.¹³ Conversely, individuals with psychopathy demonstrate impairments in emotional empathy but not cognitive empathy.¹⁴ fMRI studies showed that the affective empathy and mentalizing (a form of cognitive empathy) are independent at both behavioral and neural level.¹¹ Affective empathy primarily engages the anterior insula and anterior cingulate cortex,^3,15 regions associated with emotional processing and interoception, while cognitive empathy engages a network associated with higher-order cognitive processes of emotion regulation and perspective taking. This network primarily involves the dorsolateral and medial prefrontal cortices,^16,17,18,19 which also support executive function and cognitive control. Cognitive empathy is also known to activate areas implicated in mentalizing and theory of mind processes, such as the temporoparietal junction (TPJ), temporal pole, and precuneus/posterior cingulate cortex.^{11,20,21,22,23} Together, these findings provide evidence that the neural networks underpinning affective and cognitive empathy are distinct.

Still, a complete distinction between cognitive and affective empathy, while conceptually useful, is increasingly recognized as overly simplistic. Studies²⁴ have shown that rhythmic neural patterns during empathy for vicarious pain cannot be fully explained by this dichotomy, suggesting a more nuanced, dynamic interplay of underlying neural empathy. Recent work²⁵ further proposed a graded model of empathy and highlighted its context-dependent and flexible nature. Together, these findings challenge the strict cognitive-affective divide and support a more integrated, continuum-based conceptualization of empathy.

ERP studies of cognitive and affective empathy

Electroencephalography (EEG) studies have corroborated the intricate relationship between affective and cognitive empathy and described a complex, multi-stage process that unfolds over time and encompasses both automatic and complex cognitive responses. Research indicates that empathy follows an ontogenetic framework, beginning with primitive affective resonance that gradually matures into other-oriented responses.^26,27 This progression from simple to complex processing is evident across development; infants initially show only basic empathic responses through emotional contagion and pain attribution,^28,29 and neural responses become increasingly differentiated with age. This maturation is reflected in event-related potential (ERP) components: younger children (4–5 years) show pronounced early automatic responses peaking at 200–300 ms that diminish with age (6–9 years), suggesting better regulation of immediate affective reactions. Later cognitive components, such as the late positive potential (LPP) show expanding neural recruitment from limited activation in young children to broader engagement in older children, indicating maturation of more sophisticated cognitive processing.^30,31

The developmental progression toward increasingly complex empathic processing is mirrored in ERP studies that examined neural responses and how automatic affective reactions morph into sophisticated cognitive evaluations that unfolds over time and reflect different levels of processing.^32,33,34,35 Early automatic responses to empathy-inducing stimuli, particularly related to others’ pain, are observed as early as 100–200 ms post-stimulus onset,^31,33,36 reflecting automatic processing associated with affective empathy. Specifically, images of painful situations were found to elicit the N110 and P180 components and reflect automatic processing that represents the quick emotional resonance to others’ states.^33,34 These responses are followed by cognitive processing starting around 300–500 ms, reflecting emotional arousal and early cognitive evaluations.^35,37,38 This later phase begins to involve aspects of cognitive empathy, including initial perspective-taking and evaluation processes.³⁵

The subsequent phase, from 500 ms to approximately 700–1,000 ms post-stimulus is modulated by top-down processes that represent more deliberate, cognitive aspects of empathy.^32,33,34,35 Within this period, particularly between 450 and 650 ms, there is notable integration of affective and cognitive empathy components.³² This later phase more strongly reflects cognitive empathy processes of cognitive evaluation of the other’s situation, as evidenced by the late positive component (LPC) and sustained empathic responses.^{30,31,35,37,39,40}

During this period, cognitive processing becomes predominant, with distinct neural mechanisms emerging across different social contexts. For example, empathic responses to in-group pain engage the middle frontal gyrus (MFG), associated with direct action understanding, while out-group pain processing recruits the TPJ, involved in abstract mentalizing and self/other distinctions.⁴¹ The social context further modulates empathic processing, as evidenced by decreased LPC responses when observing individuals who committed moral transgressions compared to those exhibiting neutral behavior.³⁹ While overall engagement levels remain comparable across different social contexts during this phase, the underlying neural mechanisms reflect increasingly complex cognitive strategies. The implementation of various emotion regulation strategies during this time window modulates empathic responses,³⁰ highlighting the dynamic interplay between affective and cognitive components.

Notably, although most ERP studies focus on empathic responses within the first 1,000 ms post-stimulus, evidence suggests that evaluative and emotion-regulation processes continue well beyond this time frame. Studies examining emotional versus control stimuli have identified differential responses extending to 1,500 ms,^37,42,43 with some emotional processing persisting up to 2,000 ms or throughout stimulus presentation.^{30,37,44,45,46,47,48,49,50} This extended temporal window underscores the complex, sustained nature of empathic processing.

The neural processes occurring beyond 1,000 ms post-stimulus are predominantly associated with cognitive empathy and are hypothesized to reflect higher-order cognitive processes, such as perspective taking, cognitive emotion regulation,³⁰ cognitive reappraisal,⁵¹ and cognitive control, with reappraisal processes suggested to attenuate responses 1,500 ms following image onset.⁵⁰ Many ERP studies utilized a combination of textual and visual stimuli to evoke empathic or emotional responses and the integration of contextual information to formulate a comprehensive empathic response may also play a role in these later processing time-windows.

Notably, research targeting the empathic and cognitive aspects of empathy using ERP beyond the 1,000 ms time window remains limited. This gap in the literature presents an opportunity for our study to further explore long-latency components of empathic processing, potentially unraveling the intricate cognitive processes involved in the sustained empathic response. Moreover, by extending our investigation beyond the single-brain paradigm to examine two brains simultaneously, we can tap a novel dimension in empathy research.

Empathy to distress

Magnetoencephalography (MEG) studies have revealed distinct temporal and spatial signatures in how the brain processes empathy for physical pain versus emotional distress. Empathy for physical pain typically engages earlier sensorimotor processes, occurring within 500–900 ms after stimulus presentation.^52,53 In contrast, empathy for emotional distress involves extended processing windows (800–1,200 ms and 1,250–1,850 ms) and recruits networks associated with mentalizing and cognitive processing.^20,54 The engagement of these networks suggests that empathy to distress relies more heavily on mentalizing networks and perspective-taking processes compared to the sensorimotor focus of empathy to pain.

These investigations have delineated distinct spatiotemporal signatures for processing empathy to physical pain versus emotional distress. The earlier temporal dynamics and sensorimotor focus of pain empathy processing contrast with the later more distributed activation patterns observed during empathy for emotional distress. This dissociation between early sensorimotor resonance in pain empathy and later engagement of mentalizing networks in distress empathy aligns with theoretical models proposing distinct neural pathways for processing physical versus emotional states.

Two-brain synchrony and the empathic process

Two-brain research, a state-of-the-art approach in social neuroscience that is rooted in second-person neuroscience,⁵⁵ can address phenomena that transcend the scope of traditional single-brain studies. This approach employs hyperscanning techniques, the simultaneous recording of brain activity from multiple individuals to measure real-time neural synchrony between people during social interactions or while completing tasks. Neural synchrony in this context refers to the temporal correlation of brain activity patterns between individuals and reflects shared neural processes during social interactions or task execution.

Hyperscanning studies offer novel insights into the neural underpinnings of social processes like cognition, empathy, and interpersonal communication within naturalistic contexts.^56,57 This two-brain approach enables the investigation of interbrain dynamics during empathic processing, providing a more comprehensive understanding of empathy as it unfolds in real-world social interactions. By capturing the neural activity of both individuals simultaneously, hyperscanning reveals shared neural representations and mutual influences that facilitate empathic experiences, offering a window into the dynamic and reciprocal nature of empathic interactions.

Over the past decade, both EEG and Functional Near-Infrared Spectroscopy (fNIRS) hyperscanning showed empathy-related interbrain synchrony across participants in regions associated with social cognition and emotion processing, including the prefrontal cortex, temporal areas, and TPJ^{58,59,60,61,62}: In a dual-EEG study, interbrain synchrony emerged in temporal and central brain regions between romantic partners during naturalistic empathy-giving interactions.⁵⁹ An fNIRS hyperscanning study demonstrated increased interbrain synchrony in the right superior frontal cortex during cooperative tasks, which require empathy and perspective taking, with higher synchrony correlating with better performance.⁵⁸ In hyperscanning EEG study of empathy-to-pain, increased interbrain synchrony was observed between romantic couples and was associated with pain reduction, suggesting a potential two-brain mechanism for empathic comfort.⁶¹ Studies of mother-child pairs showed neural synchronization across development in response to empathy-eliciting and attachment-related stimuli. From early development, mothers’ and newborns’ brains show coordinated activity during pain processing, with maternal parietal cortex activity synchronizing with the infant’s somatosensory responses during heel prick procedure, as demonstrated in dual-brain fNIRS hyperscanning.⁶³ In pre-adolescence, MEG recordings show that mothers and 9-year olds demonstrate synchronized activity in the superior temporal sulcus (STS) when viewing videos of their own interactions, particularly during episodes of behavioral synchrony.⁶⁰

Collectively, these findings indicate that empathic responses are characterized by enhanced neural synchrony, reflecting shared representations and mutual understanding that are fundamental to empathy.⁶⁴ They further suggest that interbrain synchronization may serve as a neural marker for empathic processes and offer novel insights into the neural mechanisms of social and emotional contagion.^57,65 However, despite these advances, interbrain research on empathy remains in its early stages. The precise neural empathic mechanisms and their manifestation in interbrain synchrony over time have not been elucidated, and no two-brain study has systematically differentiated interbrain processes related to empathy and their unfolding in time in the absence of social interactions.

While most hyperscanning research has focused on real-time social interactions between physically co-present participants, fundamental questions remain as to the mechanisms that underpin interbrain synchronization. Traditional paradigms typically involve complex scenarios where behavioral synchrony, joint attention, and physical co-presence are intertwined, rendering it difficult to isolate their specific contributions to neural coupling. This raises the crucial question of whether interbrain synchrony exclusively depends on direct social interactions, or could it emerge from shared emotional experiences in the absence of physical co-presence? The current study takes a novel approach to address this question by deliberately separating participants while presenting identical emotional and control stimuli to assess their empathic processing. This allows us to test whether shared emotional processing alone can drive interbrain synchronization, independent of social interaction. By isolating the neural mechanisms of empathic processing from other confounding factors in interactive paradigms, we can gain new insights on the building blocks of empathic synchrony. This approach complements existing research by specifically investigating how emotional stimuli can synchronize brain responses across individuals even without direct social interaction or mutual awareness between the two.

The current study

Our study bridges two distinct approaches in empathy research: single-brain ERP studies that investigated how empathic processes unfold over time from early automatic responses to later sophisticated cognitive processing and two-brain research that evaluated neural synchrony over extended periods. By combining these approaches, we aimed to examine the temporal dynamics of interbrain empathic responses at the fine-grained temporal resolution characteristic of ERP studies.

Using hyperscanning EEG, we simultaneously recorded brain activity from two participants as they viewed empathy-to-distress and control stimuli (see Figure 1). We employed event-related techniques to assess time-locked neural responses to these stimuli across both participants. This methodology offers precise temporal resolution for investigating neural activities and enables the measurement of specific stimulus-locked responses. In the hyperscanning context, this approach allowed us to examine the coordinated brain processes between participants and track the evolution of their neural synchrony during empathic processing, from initial automatic responses through increasingly complex evaluative processes. By integrating time-locked events with hyperscanning techniques, we captured the intricate temporal dynamics of interbrain empathic responses, providing novel insights into the neural mechanisms that support empathic processing between individuals.

Figure 1.

Visualization of procedure

Mother-adolescent pairs participated in an empathy-to-distress task while seated in separate rooms, each viewing synchronized stimulus presentations. The paradigm included empathy-to-distress and control blocks (four blocks each). Each block began with a 10-s contextual sentence. For empathy blocks, sentences described emotionally challenging situations (e.g., “The child learned they must repeat the school year due to poor grades”). Control blocks presented neutral daily activities (e.g., “The child tied their shoelaces before school, a skill learned last year”). Following a fixation cross, each block presented 7–10 corresponding stimuli, displayed for 2 s with randomized 1–2 s inter-stimulus intervals. After each block, adolescents rated: (1) ease of imagining the depicted situations (5-point scale: easy to hard), and (2) perceived stress/distress levels of individuals shown (scale: 1–5). Mothers viewed their adolescent’s responses without providing ratings. This design enabled simultaneous recording of time-locked neural responses to empathy-inducing stimuli in both participants.

Taking into consideration the extended nature of neural dynamics involved in empathic processing, we assessed interbrain synchrony across four distinct time windows in a time-locked manner: 0–500 ms (early automatic responses), 500–1,000 ms (intermediate processes integrating automatic and evaluative components), 1,000–1,500 ms (complex higher-order assessments and emotional regulation), and 1500–2,000 ms (late neural responses involving reappraisal and sophisticated cognitive processing). This allowed us to track the evolution of empathic processes from rapid, automatic responses to increasingly complex evaluative processes, while assessing how these processes synchronize between individuals over time.

Our second goal was to examine whether empathic interbrain neural responses can occur in the absence of physical co-presence or social cues. Prior hyperscanning studies assessed empathic neural synchrony while individuals were in a shared physical space, leaving open the question of whether brain-to-brain alignment during empathic responses can occur when participants are physically separated. To tackle this question, we evaluated interbrain synchrony in mother-adolescent dyads when participants were seated in separate rooms. We chose to focus on mother-adolescent dyads based on previous research demonstrating their capacity to synchronize their brains even when physically separated, such as during video calls or while texting.^66,67 As both maternal and adolescent empathic abilities have been linked with social behavior^68,69,70 and oxytocin levels,⁷¹ and were found to be interconnected,⁷⁰ we measured these factors as correlates of their empathic neural alignment.

Our third goal was to test whether empathic interbrain alignment is unique to mother-child dyads. According to the biobehavioral synchrony model,^72,73 neural synchrony initially develops within the mother-infant bond and subsequently extends to other attachment relationships and broader social contexts. We investigated whether neural alignment during empathic responses remains specific to attachment partners or has evolved into a general characteristic of our species.

We proposed four hypotheses for this study. First, we predicted that mother-adolescent dyads would exhibit significant interbrain synchrony in frontotemporal regions associated with empathic responses when exposed to empathy-to-distress stimuli, compared to control stimuli (see Figure 1 for detailed paradigm). This prediction is grounded in prior research indicating that mother-child behavioral synchrony beginning in infancy provides the basis for the development of empathy in adolescence.⁷⁴ Mother-child synchrony plays a crucial role in the development of moral understanding and perspective-taking abilities, with lasting developmental effects. Supporting this, research has shown that mother-child attachment shapes neural empathic responses to others’ distress in adolescence, as measured by MEG,⁵⁴ predict the neural empathic response in young adulthood, as measure by fMRI,⁷⁵ and foreshadows interbrain synchrony in adolescence during face-to-face interactions.⁷⁶

Second, we expected the empathic neural alignment to be more pronounced in the later time windows of 1,000–2,000 ms post-stimuli. Our empathy-to-distress stimuli depicted complex emotional situations such as social isolation or personal failure, rather than physical pain. Given this complexity, we anticipated that interbrain empathic responses would primarily emerge in later time windows that capture sophisticated cognitive processes and deeper evaluation of emotional contexts. This hypothesis builds on previous MEG studies that revealed distinct temporal signatures for processing physical pain versus emotional distress.^20,52,53,54 While empathy for physical pain engages relatively early sensorimotor processes (500–900 ms),^52,53 empathy for emotional distress involves extended processing windows (800–1,200 ms and 1,250–1,850 ms) and recruits networks associated with mentalizing and cognitive integration.^20,54 Building upon this research and using a similar paradigm for investigating interbrain synchrony during empathy-to-distress, we hypothesized that empathic alignment between individuals would correspond to these later time windows, reflecting the complex nature of processing others’ emotional distress.

Third, we hypothesized that the degree of empathic neural alignment between mother-adolescent dyads would correlate with established markers of sociality. We measured both dyadic oxytocin levels and behavioral synchrony during a subsequent face-to-face interaction as potential biological and behavioral correlates of neural synchronization during empathy stimulus processing. This approach was guided by oxytocin’s well-documented role in social bonding and empathy,^77,78 and by previous research showing that mother-child oxytocin levels are interrelated and linked to empathic abilities.^70,71 Similarly, dyadic behavioral synchrony, the coordinated social behavior between mother and child—represents a foundational aspect of attachment relationships that shapes social development. Maternal behaviors have been found to predict adolescents’ neural responses to others’ distress,⁵⁴ while mother-child behavioral synchrony significantly influences children’s neural development in social-cognitive and empathy-related circuits.^75,79,80 By examining both biological (oxytocin) and behavioral (dyadic behavioral synchrony in a later interaction) markers alongside neural synchronization during empathy processing, we aimed to establish whether interbrain coupling during empathic processing reflects broader patterns of biological and behavioral attunement within the mother-adolescent relationship.

Finally, we investigated whether unacquainted individuals would exhibit empathic neural alignment, an unexplored question in current research, for lack of prior research. The presence of similar interbrain synchrony patterns between strangers as those observed in mother-adolescent dyads would suggest a universal neural mechanism for empathic alignment in our species.³ However, the absence of social cues and pre-existing relationships might impede strangers’ ability to synchronize during empathy-inducing situations, particularly when physically separated. This exploratory component addresses the role of relationship closeness and familiarity in neural empathic alignment, potentially expanding our understanding of the neural mechanisms underlying empathy and social cognition.

Results

Study 1

Comparing interbrain synchrony of mother-adolescent dyads during empathy-to-distress vs. control stimuli processing

To examine stimulus-specific effects of empathy-to-distress and control stimuli on interbrain synchrony, we analyzed weighted phase lag index (wPLI) values across the frontotemporal network between spatially separated mother-adolescent dyads. The analysis evaluated neural synchronization during both empathy-to-distress and control conditions, focusing on two key frequency bands: alpha (8–13 Hz) and beta (13.5–29.5 Hz). To characterize the temporal dynamics of neural responses, we divided the 2,000 ms stimulus presentation period into four consecutive 500 ms time windows. For each frequency band, we conducted repeated measures ANOVA with two within-subject factors: stimulus type (empathy-to-distress vs. control) and time window (0–500, 500–1,000, 1,000–1,500, and 1,500–2,000 ms post-stimulus onset). This analysis allowed us to track how interbrain synchrony evolved over time and differed between empathy-inducing and control conditions.

Alpha

For the alpha band (8–13 Hz), repeated measures ANOVA revealed a significant main effect of stimulus type (F(1, 49) = 4.03, p = 0.05, η²p = 0.08), with empathy-to-distress stimuli eliciting greater interbrain synchrony than control stimuli. Neither the main effect of time window (F(3, 147) = 2.03, p = 0.112, η²p = 0.04) nor the interaction between stimulus type and time window (F(3, 147) = 0.75, p = 0.52, η²p = 0.02) reached significance, indicating that enhanced synchrony for empathy-to-distress stimuli remained consistent throughout the entire stimulus presentation period.

Beta

Analysis of the beta band (13.5–29.5 Hz) similarly showed a significant main effect of stimulus type (F(1, 49) = 5.21, p = 0.027, η²p = 0.026), with greater synchrony for empathy-to-distress stimuli. While no main effect of Time window emerged (F(3, 147) = 0.35, p = 0.79, η²p = 0.003), a significant interaction of stimulus type and time window was found (F(3, 147) = 3.28, p = 0.023, η²p = 0.024), indicating that differences between stimulus types varied across time windows in this frequency band.

To examine the interaction effect, we conducted false discovery rate (FDR)-corrected post-hoc paired-samples t tests comparing interbrain synchrony between empathy-to-distress and control stimuli within each time window. Analysis of the early time windows revealed no significant increases in interbrain synchrony for empathy-to-distress stimuli during either 0–500 ms (t(49) = 0.551, p_{(FDR-corrected)} = 0.584, Cohen’s d = 0.078) or 500–1,000 ms time windows (t(49) = 1.12, p_{(FDR-corrected)} = 0.36, Cohen’s d = 0.16). However, a significant enhancement in interbrain synchrony emerged during the 1,000–1,500 ms window (t(49) = 3.33, p_{(FDR-corrected)} = 0.008, Cohen’s d = 0.47). This effect diminished in the final 1,500–2,000 ms window, where the difference did not maintain statistical significance after FDR correction (t(49) = 1.95, p_{(FDR-corrected)} = 0.114, Cohen’s d = 0.28) (see Table 1; Figure 2). These findings demonstrate that enhanced interbrain synchrony during empathy-to-distress stimuli emerges specifically during the 1,000–1,500 ms time window, suggesting that neural alignment between mother and child peaks during the period associated with higher-order cognitive processing of empathic responses.

Table 1.

Mean (SD) of interbrain synchrony (wPLI) while viewing control vs. empathy-to-distress stimuli

Time window	Control (SD)	Empathy (SD)	t-value	p(FDR-corrected)	Cohen’s d
Alpha
0–500 ms	0.216 (0.026)	0.218 (0.029)	0.414	0.48	0.059
500–1,000 ms	0.208 (0.025)	0.211 (0.023)	0.68	0.67	0.096
1,000–1,500 ms	0.209 (0.023)	0.214 (0.024)	1.2	0.68	0.17
1,500–2,000 ms	0.209 (0.021)	0.219 (0.031)	2.264	0.11	0.32
Beta
0–500 ms	0.194 (0.016)	0.192 (0.024)	0.551	0.584	0.078
500–1,000 ms	0.192 (0.017)	0.196 (0.023)	1.116	0.36	0.158
1,000–1,500 ms	0.188 (0.016)	0.199 (0.020)	3.325	0.008 ∗∗	0.47
1,500–2,000 ms	0.192 (0.017)	0.198 (0.023)	1.95	0.114	0.275

Temporal dynamics of interbrain synchrony during empathic processing. Mean interbrain synchrony values comparing empathy-to-distress versus control stimuli across four sequential 500 ms time windows. Both alpha (8–13 Hz) and beta (13.5–29.5 Hz) bands showed enhanced synchrony during empathy-to-distress stimuli across the full 2,000 ms period. However, only the beta band exhibited a significant interaction between Stimulus type and Time window interaction, characterized by greater interbrain synchrony during later processing stages (1,000–1,500 ms) ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Figure 2.

Visualization of frontotemporal neural synchrony in alpha (8–13 Hz) and beta (13.5–29.5 Hz) bands during empathy-to-distress processing in mother-adolescent dyads

Visualization of the interbrain connectivity values (weighted phase lag index; wPLI) measured across the 2,000 ms stimulus presentation interval. Left: bar graphs show mean connectivity values for empathy-to-distress and control stimuli in the alpha band (8–13 Hz). Right: bar graphs show mean connectivity values in the beta band (13.5–29.5 Hz). Error bars represent standard error of the mean (SEM). Individual data points are overlaid to show the distribution of connectivity values across mother-adolescent dyads. (∗p < 0.05).

Evaluating the contributions of ROIs to interbrain synchrony

To examine potential regional variations in neural coupling, we analyzed interbrain synchrony across all 16 possible region of interest (ROI) combinations within the frontotemporal network. For each frequency band, we conducted a repeated measures ANOVA with two within-subject factors: ROI combination (16 levels) and stimulus type (empathy-to-distress vs. control). This analysis focused on the complete 0–2,000 ms stimulus presentation window, which showed significant effects in both alpha and beta frequency bands.

Alpha

In the alpha band (8–13 Hz), the analysis confirmed the main effect of stimulus type (F(1, 49) = 4.03, p = 0.05, η²p = 0.076), with empathy-to-distress stimuli eliciting enhanced interbrain synchrony compared to control stimuli. However, neither a main effect of ROI (F(15, 735) = 0.61, p = 0.87, η²p = 0.012) nor an interaction between stimulus type and ROI (F(15, 735) = 0.802, p = 0.68, η²p = 0.016) reached significance. These results indicate that enhanced synchrony during empathy processing was distributed across the frontotemporal network rather than localized to specific regional connections.

Beta

Analysis of the beta band revealed a significant main effect of stimulus type (F(1, 49) = 5.21, p = 0.027, η²p = 0.096), with empathy-to-distress stimuli eliciting enhanced interbrain synchrony compared to control stimuli. No significant effects were observed for ROI (F(15, 735) = 1.19, p = 0.27, η²p = 0.024) or the interaction between stimulus type and ROI (F(15, 735) = 0.79, p = 0.69, η²p = 0.016). These findings indicate that enhanced neural synchronization during empathic processing was distributed across the frontotemporal network rather than being confined to specific regional connections. Connectivity values for all ROI combinations in both frequency bands are presented in Table 2.

Table 2.

Mean (SD) of interbrain synchrony (wPLI) for each ROI while viewing control vs. empathy-to-distress stimuli

ROI	Alpha empathy (mean ± SD)	Alpha control (mean ± SD)	Beta empathy (mean ± SD)	Beta control (mean ± SD)
LF-LF	0.211 (0.031)	0.210 (0.035)	0.190 (0.026)	0.191 (0.024)
LF-LT	0.215 (0.030)	0.207 (0.023)	0.198 (0.030)	0.192 (0.023)
LF-RF	0.214 (0.026)	0.209 (0.027)	0.193 (0.027)	0.189 (0.020)
LF-RT	0.219 (0.029)	0.211 (0.035)	0.193 (0.026)	0.194 (0.021)
LT-LF	0.216 (0.031)	0.205 (0.023)	0.199 (0.028)	0.194 (0.029)
LT-LT	0.215 (0.037)	0.213 (0.030)	0.199 (0.029)	0.191 (0.025)
LT-RF	0.217 (0.032)	0.213 (0.029)	0.198 (0.027)	0.188 (0.022)
LT-RT	0.216 (0.030)	0.209 (0.027)	0.202 (0.024)	0.192 (0.026)
RF-LF	0.214 (0.037)	0.211 (0.030)	0.197 (0.028)	0.195 (0.032)
RF-LT	0.220 (0.034)	0.205 (0.030)	0.193 (0.025)	0.189 (0.022)
RF-RF	0.214 (0.033)	0.209 (0.027)	0.197 (0.026)	0.187 (0.022)
RF-RT	0.216 (0.035)	0.207 (0.025)	0.198 (0.026)	0.189 (0.022)
RT-LF	0.212 (0.036)	0.215 (0.030)	0.197 (0.025)	0.191 (0.022)
RT-LT	0.212 (0.032)	0.213 (0.029)	0.197 (0.030)	0.196 (0.027)
RT-RF	0.218 (0.040)	0.212 (0.033)	0.192 (0.027)	0.187 (0.027)
RT-RT	0.219 (0.035)	0.218 (0.032)	0.199 (0.032)	0.196 (0.025)

Interbrain synchrony values for each ROI during the 0–2,000 ms stimulus presentation window. Mean (SD) interbrain synchrony values for each ROI combination across the 2,000 ms stimulus presentation window in alpha (8–13 Hz) and beta (13.5–29.5 Hz) frequency bands. While enhanced synchrony was observed for empathy-to-distress compared to control stimuli, no significant ROI effects (all p > 0.25) or stimulus type × ROI interactions were found in either frequency band.

Bio-behavioral correlates of neural synchrony

To investigate potential mechanisms underlying enhanced neural coupling during empathic processing, we examined the associations between the increase in interbrain synchrony during empathic processing and key bio-behavioral measures. Given the more pronounced effects observed in the beta band, these analyses focused on beta-band neural synchrony across the full 2,000 ms stimulus presentation period, comparing empathy-to-distress versus control conditions. We specifically investigated two factors: baseline oxytocin levels as a neuroendocrine marker and dyadic behavioral synchrony during a subsequent face-to-face interaction as a measure of dyadic social coordination.

Association between oxytocin levels and interbrain synchrony during empathic processing

To examine the relationship between neuroendocrine function and neural coupling, we analyzed the association between baseline oxytocin (OT) levels and stimulus-specific enhancement of interbrain synchrony. Baseline OT levels were collected from both mothers and adolescents prior to the empathy task. Initial analysis revealed a marginally significant correlation between mother and child OT levels (r = 0.26, p = 0.077), suggesting potential coordination in baseline neuroendocrine states within dyads. Analysis of dyadic OT (calculated as the sum of mother and child OT levels) revealed a significant positive correlation with the magnitude of enhanced interbrain synchrony during empathy-to-distress stimuli relative to Control stimuli across the 2,000 ms presentation period (r = 0.31, p = 0.036; Figure 3A), suggesting that baseline neuroendocrine state may influence the degree of neural coupling during empathic processing.

Figure 3.

Biobehavioral factors predicting enhanced interbrain synchrony during empathic processing

Scatter plots depicting the relationships between biobehavioral measures and neural synchrony enhancement (calculated as the difference between empathy-to-distress and control conditions). Each point represents one dyad. Dark gray shaded area indicates 95% confidence interval around the regression line; light gray shaded area indicates 95% prediction interval.

(A) Positive correlation between dyadic oxytocin levels and increased interbrain synchrony during empathic processing.

(B) Positive correlation between behavioral synchrony scores (measured during subsequent face-to-face interaction) and increased interbrain synchrony during empathic processing. (∗p < 0.05 and ∗∗p < 0.01).

Brain-behavior coupling: Behavioral synchrony predicts neural alignment

To investigate the relationship between neural and behavioral measures of dyadic behavioral synchrony, we examined the associations between the enhancement in interbrain synchrony during empathic processing and behavioral synchrony assessed during a subsequent face-to-face interaction. Behavioral synchrony was quantified using the coding interactive behavior (CIB) system, focusing on the key components of dyadic synchronous behavior: dyadic reciprocity, dyadic flow, and dyadic relaxed behavior (reverse-coded from dyadic tension).

Correlation analysis revealed a significant positive association between dyadic behavioral synchrony scores and the enhancement of neural synchrony during empathy-to-distress stimuli compared to control stimuli (r = 0.41, p = 0.004; Figure 3B). This correlation indicates that dyads exhibiting stronger behavioral synchrony during face-to-face interactions also demonstrated enhanced neural coupling during empathic processing, despite the absence of physical co-presence during the empathy task.

Participants’ subjective assessment of stimulus content

Our following analysis examined the participants’ subjective ratings of both empathy-to-distress and control stimuli, collected at the conclusion of each experimental block. These ratings assessed two key dimensions: the perceived level of distress of the depicted individuals and the participants’ ability to mentally simulate the presented situations.

For perceived distress levels, the Shapiro-Wilk test indicated non-normal data distribution (W = 0.938, p = 0.020), prompting the use of a Wilcoxon signed-rank test. The results revealed that participants rated individuals in the empathy-to-distress stimuli as significantly more distressed (M = 4.409) compared to the control stimuli (M = 1.795), z = −5.777, p < 0.001.

Regarding ease of imagining the situation of the individuals depicted in the stimuli, the Shapiro-Wilk test suggested marginal deviations from normality (W = 0.950, p = 0.055). A Wilcoxon signed-rank test showed that participants found the empathy-to-distress condition scenarios as harder to imagine (M = 2.847) compared to control scenarios (M = 1.852), z = −4.295, p < 0.001.

Study 2

Interbrain synchrony in stranger dyads: Empathy-to-distress vs. control conditions

To investigate whether the neural coupling patterns observed in mother-adolescent dyads generalize to unacquainted individuals, we examined interbrain synchrony between paired strangers with no prior interaction. Following the analytical approach of study 1, we calculated weighted phase lag index (wPLI) values across the frontotemporal network during both empathy-to-distress and control conditions. The analysis examined neural synchronization across four consecutive 500 ms time windows spanning the complete 2,000 ms stimulus presentation period.

Alpha

For the alpha band (8–13 Hz), an analysis using repeated measures ANOVA with stimulus type (empathy-to-distress vs. control) and time window (0–500, 500–1,000, 1,000–1,500, and 1,500–2,000 ms post-stimulus) revealed a significant main effect of time window (F(3, 63) = 3.82, p = 0.014, η²p = 0.15). No significant effect was observed for stimulus type (F(1, 21) = 0.278, p = 0.604, η²p = 0.01) or an interaction between stimulus type and time window (F(3, 63) = 1.55, p = 0.21, η²p = 0.07).

Follow-up analyses of the time window effect using FDR-corrected t tests showed marginally greater synchrony in the earliest time window (0–500 ms) compared to both the 500–1,000 ms window (t(21) = 2.78, p(FDR-corrected) = 0.06) and 1,500–2,000 ms window (t(21) = 2.51, p(FDR-corrected) = 0.06). No significant differences were observed between other time windows: 0–500 vs. 1,000–1,500 ms (t(21) = 1.80, p(FDR-corrected) = 0.18), 500–1,000 vs. 1,000–1,500 ms (t(21) = 1.59, p(FDR-corrected) = 0.195), 500–1,000 vs. 1,500–2,000 ms (t(21) = 1.23, p(FDR-corrected) = 0.28), and 1,000–1,500 vs. 1,500–2,000 ms (t(21) = 0.2, p(FDR-corrected) = 0.84). This pattern suggests an early peak in synchronization that gradually diminishes over time, regardless of stimulus type.

Beta

Building upon the results of study 1, which revealed greater interbrain synchrony for mother-adolescent dyads during a specific time window of empathic processing, our initial analysis for the stranger dyads study focused on the 1,000–1,500 time window, which have been previously found to exhibit greater interbrain synchrony for empathy-to-distress stimuli compared to control stimuli.

We first conducted a priori paired-samples t test to compare interbrain synchrony between empathy-to-distress and control conditions during the 1,000–1,500 ms window. Results revealed significantly increased synchrony (t(21) = 2.18, p = 0.041, Cohen’s d = 0.46), indicating that similar to mother-adolescent dyads, stranger dyads exhibited similar enhancement of neural synchrony during this phase of empathic processing (see Figure 4).

Figure 4.

Comparison of interbrain synchrony between mother-adolescent and stranger dyads

Interbrain connectivity (weighted phase lag index; wPLI) during the 1,000–1,500 ms post-stimulus window. Error bars represent standard error of the mean (SEM). Individual data points are overlaid to show distribution across dyads. Left: bar graphs show mean connectivity values for mother-adolescent dyads during empathy-to-distress versus control stimuli. Right: corresponding connectivity values for stranger dyads. Both groups demonstrated significantly enhanced synchrony during empathy-to-distress stimuli (p < 0.025), and no differences between the groups were observed, suggesting a shared neural mechanism for empathic processing independent of relationship type.

To examine potential differences from study 1, we next conducted post-hoc analyses of other time windows. The results indicated that no significant increases in synchrony were observed during the earlier processing phases (0–500 ms: t(21) = 0.23, p = 0.99, Cohen’s d = 0.048; 500–1,000 ms: t(21) = 0.013, p = 0.99, Cohen’s d = −0.003) or in the late processing phase (1,500–2,000 ms: t(21) = 0.05, p = 0.99, Cohen’s d = 0.011; see Table 3).

Table 3.

Mean (SD) of interbrain synchrony (wPLI) while viewing control vs. empathy-to-distress stimuli

Time window	Control (SD)	Empathy (SD)	t	p(FDR-corrected)	Cohen’s d
Alpha
0–500 ms	0.23 (0.036)	0.244 (0.045)	1.167	0.256	0.25
500–1000 ms	0.222 (0.036)	0.213 (0.025)	0.097	0.341	0.21
1000–1500 ms	0.228 (0.031)	0.224 (0.025)	0.558	0.582	0.12
1500–2000 ms	0.22 (0.037)	0.23 (0.037)	1.34	0.196	0.29
Beta
0–500 ms	0.198 (0.022)	0.199 (0.022)	0.227	0.823	0.048
500–1000 ms	0.193 (0.024)	0.193 (0.021)	0.013	0.990	0.003
1000–1500 ms	0.194 (0.024)	0.205 (0.021)	2.177	0.041∗	0.464
150–02000 ms	0.196 (0.017)	0.196 (0.028)	0.053	0.958	0.011

Time-resolved analysis of interbrain synchrony during empathic processing. Weighted phase lag index (wPLI) values comparing empathy-to-distress versus control stimuli across sequential 500 ms time windows. Enhanced synchrony was specifically observed in the beta band (13.5–29.5 Hz) during the 1,000–1,500 ms post-stimulus window, coinciding with higher-order cognitive processing of empathic responses. Mean (SD) values are presented for both stimulus conditions. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Evaluating the contributions of ROIs to interbrain synchrony

To examine the spatial specificity of neural coupling, we analyzed interbrain synchrony across all 16 ROI combinations within the frontotemporal network. Given that stimulus-specific effects were observed exclusively in the beta frequency band during the 1,000–1,500 ms time window, we focused our regional analysis on this specific time-frequency window. We conducted repeated measures ANOVA with two within-subject factors: ROI combination (16 levels) and stimulus type (empathy-to-distress vs. control).

The analysis confirmed the main effect of stimulus type (F(1, 21) = 4.74, p = 0.041, η²p = 0.18), with empathy-to-distress stimuli eliciting enhanced synchrony compared to control stimuli. However, neither an effect of ROI (F(15, 315) = 1.142, p = 0.32, η²p = 0.05) nor an interaction between stimulus type and ROI (F(15, 315) = 0.95, p = 0.52, η²p = 0.043) reached significance, suggesting processing was distributed across the measured network rather than localized to specific regional connections.

Comparative analysis of interbrain synchrony: Mother-adolescent vs. stranger dyads

To evaluate potential differences in neural coupling patterns between attachment partners and unacquainted individuals, we conducted a direct comparison of interbrain synchrony between studies 1 and 2. We performed a mixed ANOVA with sample (mother-adolescent vs. stranger dyads) as a between-subjects factor and two within-subject factors: stimulus type (empathy-to-distress vs. control) and time window (0–500, 500–1,000, 1,000–1,500, and 1,500–2,000 ms post-stimulus).

Alpha

Analysis of the alpha-band synchrony revealed a significant main effect of sample (F(1, 70) = 9.36, p = 0.003, η²p = 0.12), indicating distinct patterns of neural coupling between mother-adolescent and stranger dyads. Given these fundamental differences in synchronization patterns, we analyzed the samples separately (see sections “comparing interbrain synchrony of mother-adolescent dyads during empathy-to-distress vs. control stimuli processing” and “interbrain synchrony in stranger dyads: Empathy-to-distress vs. control conditions” for mother-adolescent and stranger dyads, respectively). These findings suggest that relationship type substantially influences neural system alignment in the alpha rhythm during empathic processing.

Beta

Analysis of beta-band synchrony revealed no significant differences between mother-adolescent and stranger dyads (F(1, 70) = 0.57, p = 0.45, η²p = 0.01). Given this similarity, we combined the samples for subsequent analyses examining interbrain synchrony across stimulus types and time windows. While we observed no main effect of time window (F(3, 213) = 0.34, p = 0.8, η²p = 0.005), there was a significant main effect of stimulus type, with enhanced interbrain synchrony during empathy-to-distress compared to control stimuli (F(1, 71) = 5.93, p = 0.017, η²p = 0.077). Additionally, we found a significant interaction between stimulus type and time window (F(3, 213) = 4.06, p = 0.008, η²p = 0.05), indicating that the pattern of interbrain synchrony varied across time windows.

To examine this interaction, we conducted FDR-corrected post-hoc analyses for each time window, following the analytical approach used in study 1. We employed paired-samples t tests, corrected for multiple comparisons, to assess differences in interbrain synchrony between empathy-to-distress and control stimuli within each time window. The results indicate no significant difference in interbrain synchrony for the empathy-to-distress relative to control stimuli in the 0–500 ms (t(71) = −0.35, p_{(FDR-corrected)} = 0.73, Cohen’s d = 0.04), the 500–1,000 ms (t(71) = 0.87, p_{(FDR-corrected)} = 0.47, Cohen’s d = 0.11) or the 1,500–2,000 ms (t(71) = 1.69, p_{(FDR-corrected)} = 0.23, Cohen’s d = 0.19) time windows. However, a significant increase in interbrain synchrony for empathy-to-distress stimuli was observed in the 1,000–1,500 ms time window (t(71) = 4.03, p_{(FDR-corrected)} < 0.001, Cohen’s d = 0.47). Together, these results suggest that a time-sensitive, shared human mechanism of empathy is evident, reflecting neural alignment of higher-order cognitive processing of empathic responses that peak in the 1,000–1,500 time window and then wanes in the following 1,500–2,000 time window.

Participants’ subjective assessment of stimulus content

Our next analysis examined participants’ ratings of the empathy-to-distress versus control stimuli. We evaluated the two key aspects that participants were asked to rate at the conclusion of each block: the perceived distress level of the individuals depicted in the stimuli, and participants’ ability to imagine the depicted situations.

For perceived distress levels, the Shapiro-Wilk test indicated non-normal data distribution (W = 0.863, p = 0.006), prompting the use of a Wilcoxon signed-rank test. Results revealed that participants rated individuals in the empathy-to-distress stimuli as significantly more distressed (M = 4.5) compared to the control stimuli (M = 1.470), z = 4.107, p < 0.001.

Regarding ease of imagining the situation of the individuals depicted in the stimuli, the Shapiro-Wilk test suggested marginal deviations from normality (W = 0.924, p = 0.092). A Wilcoxon signed-rank test showed that participants did not find the empathy-to-distress scenarios (M = 1.64) as different to imagine compared to control scenarios (M = 1.360), z = −1.533, p = 0.130.

Comparison of participants’ ratings across both studies

Ease of situation visualization

A mixed ANOVA has been conducted to examine how the experimental condition (empathy-to-distress vs. control) and sample type (mother-adolescent vs. strangers) affected participants’ ratings of imaginability for the depicted situations.

The analysis revealed a significant main effect of condition (F(1,64) = 19.398, p < 0.001, η² = 0.23), with participants rating situations in the empathy-to-distress condition (M = 2.247, SD = 1.107) as harder to imagine than those in the control condition (M = 1.606, SD = 0.647). Additionally, a significant effect of sample was also observed (F(1,64) = 27.461, p < 0.001, η² = 0.3), with the mother-adolescent sample reporting greater difficulty imagining situations across both conditions (M = 2.260, SD = 0.774) compared to the stranger sample (M = 1.852, SD = 0.699).

Notably, these main effects were qualified by a significant interaction between condition and sample, (F(1,64) = 6.19, p = 0.015, η² = 0.09). As observed in the individual ratings of each study, in the mother-adolescent sample, participants rated empathy-to-distress situations (M = 2.847, SD = 1.107) as significantly more difficult to imagine than control situations (M = 1.360, SD = 0.395). In contrast, the stranger sample showed comparable imaginability ratings between empathy-to-distress (M = 1.636, SD = 0.743) and control conditions (M = 1.852, SD = 0.699).

Perceived distress ratings

We conducted a mixed ANOVA to examine how experimental condition (empathy-to-distress vs. control) and sample type (mother-adolescent vs. strangers) influenced participants’ ratings of situation imaginability.

Analysis of within-subjects effects revealed a significant main effect of condition, (F(1,64) = 1499.56, p < 0.001, η² = 0.96), with participants rating the distress level of the individuals in the empathy-to-distress stimuli (M = 4.455, SD = 0.423) as significantly higher compared to the control stimuli (M = 1.633, SD = 0.477). A between-subjects analysis showed no significant main effect of sample type (F(1,64) = 1.462, p = 0.231, η² = 0.02), indicating that overall, the perceived stress level ratings did not differ between the two groups.

Finally, a significant interaction of Condition and Sample was found (F(1,64) = 8.173, p = 0.006, η² = 0.11). While both groups rated empathy-to-distress stimuli as more distressing than control stimuli, this effect was more pronounced in the stranger sample (empathy-to-distress: M = 4.500, SD = 0.423; control: M = 1.470, SD = 0.457) compared to the mother-adolescent sample (empathy-to-distress: M = 4.409, SD = 0.476; control: M = 1.795, SD = 0.457; see Table 4).

Table 4.

Mean (SD) of participants’ ratings while viewing control vs. empathy-to-distress stimuli

	Difficulty to imagine		Perceived stress
	Control	Empathy	Control	Empathy
Mother-adolescent dyads	1.85 (0.7)	2.85 (1.11)	1.795 (0.48)	4.41 (0.48)
Strangers dyads	1.36 (0.4)	1.64 (0.74)	1.47 (0.46)	4.5 (0.42)

Detailed account of the participants’ mean ratings of empathy-to-distress stimuli relative to control stimuli. Reported here are participants’ ratings, as a factor of sample (mother-adolescent, stranger dyads) and stimulus type (empathy-to-distress, control). The results indicate that empathy-to-distress stimuli were perceived as depicting greater distress relative to the control stimuli. Furthermore, in the mother-adolescent sample, participants rated the empathy-to-distress stimuli more difficult to imagine than the control condition. However, the strangers sample showed no significant difference in imaginability ratings between empathy-to-distress and control conditions. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

Discussion

Empathy and its underlying neural mechanisms play a crucial role in human social life by enabling individuals to understand and respond to others’ emotional states.^3,5 While evidence supports the importance of biobehavioral synchrony in social-cognitive processes,^56,72,73,81 the temporal dynamics and specificity of neural coupling during empathic responses remain poorly understood. The present study addressed these gaps and examined how empathic processes manifest in interbrain synchronization between individuals, in the absence of co-presence.

Our study takes a comprehensive approach to the role of interbrain synchrony in affective and cognitive empathic responses. First, we assessed the temporal dynamics underpinning interbrain empathic responses and the time resolution of these complex neural processes by utilizing a time-locked event technique during an empathy-to-distress task to differentiate automatic, intermediate, and higher-order processes. Second, we examined whether the neural alignment of empathic responses is possible when participants are not co-present and in the absence of social cues, such as speech, facial expressions, body language, or chemosignals that facilitate brain synchronization. Third, we looked at the specificity of the interbrain neural empathic response mechanism, whether it is unique to dyads sharing an affiliative bond or whether empathic neural alignment is a universal mechanism. This comparative approach allowed us to evaluate the generalizability of interbrain neural empathy beyond close relationships.

Several important insights are revealed by our findings and may advance our understanding of the neural substrates of human empathy. Consistent with our first hypothesis, exposure to empathy-eliciting stimuli enhanced neural synchronization within the frontotemporal network between both mother-adolescent dyads and strangers even without social interaction. This demonstrates, for the first time, that empathy aligns brains in synchrony even in the absence of physical co-presence and suggests the existence of an interbrain shared mechanism underpinning human empathy.⁵⁶ Consistent with the literature on the association between oxytocin and empathy,^77,78 we found that higher dyadic oxytocin levels correlated with greater increase in interbrain empathic response. Similarly, greater behavioral synchrony between mother and adolescent also linked with greater neural empathic alignment, extending prior studies to the two-brain domain.^{66,76,82,83,84,85}

Study 2 investigated whether the mechanism of interbrain empathic alignment extends beyond attachment partners to unfamiliar individuals. Our findings revealed that while both attachment partners and strangers exhibited increased neural coupling during empathy processing, distinct patterns of alignment emerged between groups. Mother-adolescent dyads exhibited significantly more robust and sustained synchronization, characterized by enhanced interbrain coherence across both alpha and beta rhythms throughout the entire stimulus presentation window. In contrast, stranger dyads showed a more temporally constrained pattern of synchronization, with increased interbrain coherence limited to the beta band during a specific time window of 1,000–1,500 ms post-stimulus onset for empathy-to-distress stimuli.

These findings advance our understanding of empathy’s neural foundations in several key ways. First, these results support the hypothesis of a universal shared neural mechanism underlying empathic responses in humans and align with established theories on shared representations in social cognition.^86,87,88 Second, they suggest a hierarchical organization of empathic neural coupling, where basic synchronization mechanisms are present across all human interactions but are significantly enhanced within established attachment relationships. The quality of this synchronization—specifically its strength and temporal dynamics—appears to be modulated by relationship status, with mother-adolescent dyads showing more robust and sustained patterns of synchronization. Third, our results demonstrate how biological and social factors interact to shape empathic processing, suggesting that the relationship within attachment relationships, particularly mother-child bonds, may create neural channels for empathic resonance that transcend immediate social context. This finding aligns with previous research establishing the fundamental role of maternal care in shaping children’s social behavior and empathic abilities.^75,79,80

This integrated perspective provides new insights into how the human brain balances universal empathic capabilities with relationship-specific enhancements. By demonstrating both the universality of neural synchronization mechanisms and their relationship-dependent modulation, our findings contribute to a more nuanced understanding of the neural architecture underlying social cognition and empathic processing across different interpersonal contexts.

Notably, compelling evidence for the importance of this mechanism comes from our analysis of participants' subjective ratings and neural synchronization patterns. Despite stranger dyads reporting greater ease in imagining the presented scenarios, they exhibited significantly less neural synchrony compared to mother-adolescent pairs in the alpha band, and more limited synchrony in the beta band. This dissociation between subjective experience and neural synchronization suggests that the enhanced neural coupling observed in mother-adolescent dyads is primarily driven by their unique attachment relationship rather than stimulus accessibility or processing ease.

A key methodological innovation in our study was the use of time-locked events to examine the temporal dynamics of interbrain synchrony during specific windows of empathic processing. This approach revealed distinct synchronization patterns: mother-adolescent pairs showed enhanced synchronization in both alpha and beta bands throughout the response, while strangers exhibited synchronization only in the beta band during the 1,000–1,500 ms post-stimulus window. This temporal profile aligns with current ERP models of empathic processing as a multi-stage phenomenon, progressing from rapid, automatic responses to more deliberate cognitive evaluations.^32,33,34,35

Our findings complement and extend previous research on the temporal dynamics of empathy processing, particularly in differentiating responses to physical versus emotional distress.^20,52,53,54 Previous MEG studies have documented a developmental progression in pain empathy processing: children primarily show alpha enhancement in primary somatosensory cortex and central sulcus, adolescents display combined alpha-beta patterns in bilateral parietal lobe and middle cingulate cortex (MCC), while adults exhibit the most sophisticated response, involving alpha suppression, beta suppression, and gamma enhancement in visceromotor cortex.⁵² Recent studies have further identified alpha-band suppression during 500–900 ms in right inferior temporal cortex during pain empathy processing.⁵³ This developmental trajectory suggests a developmental trajectory from basic sensorimotor resonance to complex interoceptive processing.

In contrast, the neural dynamics of emotional distress empathy reveal a temporally organized process engaging distinct neural networks. The response unfolds in a sequential pattern, beginning with early alpha suppression (0–850 ms), followed by alpha-band activity (7–12 Hz) in the supplementary motor area (SMA) and MCC during 800–1,200 ms, and culminating in a later processing stage (1,250–1,850 ms).^20,54 These temporal stages are associated with distinct neural sources: the TPJ and posterior STS are particularly active during later processing stages, while the MCC and SMA play crucial roles in integrating cognitive and affective components throughout the response. This temporal and spatial organization suggests that empathy for emotional distress relies heavily on mentalizing networks and perspective-taking processes, distinguishing it from the more sensorimotor-focused nature of pain empathy.

The observed interbrain synchronization in this later time window therefore likely reflects the shared recruitment of higher-order cognitive processes required for understanding and resonating with others’ emotional states, rather than basic sensorimotor resonance typically seen in response to physical pain. Additionally, distress empathy engages alpha-band activity (7–12 Hz) during 800–1,200 ms in the supplementary motor area (SMA) and MCC, regions that support integration of cognitive and affective components of empathy.⁵⁴ Our results add to these findings, and show for the first time, a shared neural mechanism between individuals underpinning empathy-to-distress that extends beyond automatic processing, transcends the single brain, and aligns brains in synchrony during empathic processing. These findings, and specifically the increase in interbrain synchrony at the 1,000 ms time-window following stimulus onset, are also consistent with previous ERP research on the processing of emotional or empathy-related components persisting beyond 1,000 ms post-stimulus and potentially continuing for the entire duration of stimuli presentation even up to 6,000 ms.^{30,37,44,45,46,47,48,49,50}

The temporal dynamics observed in our study correspond to the later stages of empathic processing identified in previous research. While mother-adolescent dyads showed sustained neural alignment throughout the task in both alpha and beta bands, in the beta band a particularly strong synchronization emerged specifically during the 1,000–1,500 ms window, driving the interaction between relationship type and neural synchronization. Stranger dyads exhibited similar synchronization exclusively during this same time window and only in the beta band. The temporal specificity corresponds to later stages of empathic processing identified in previous research,^20,54 suggesting shared neural processes associated with complex cognitive evaluations, including understanding others’ affect, self-other differentiation, formulation of helping behaviors, and emotion regulation strategies.^3,9,87,89 Notably, no interbrain synchrony was detected specifically for the automatic components of affective empathy relative to control stimuli, consistent with the nature of our experimental task and stimuli. Our study utilized stimuli predominantly reflecting complex distress that stems from social isolation, loneliness, or feelings of personal failure, rather than physical pain. The sustained synchronization in attachment pairs across frequency bands, with enhanced beta-band coupling during the 1,000–1,500 ms window, may reflect their enhanced capacity for shared emotional processing, while the more temporally constrained synchronization in stranger dyads suggests a more deliberate, cognitive engagement with the stimuli. Previous research has shown that stimuli associated with physical pain typically elicit more immediate activation of processes associated with embodiment and pain response,^{20,32,33,34,35,52,90} relative to empathy-to-distress stimuli^53,54 and future studies may assess whether pain stimuli would trigger different patterns of interbrain synchrony, particularly at earlier time-windows.

Our findings contribute to a more nuanced understanding of the temporal dynamics of empathic responses, particularly in the context of complex social and emotional situations.^89,91 The differential patterns of synchronization between attachment partners and strangers suggest that relationship status not only modulates the strength of neural coupling but also its temporal characteristics, with established relationships facilitating more sustained and comprehensive empathic processing across frequency bands.

Interbrain synchrony was assessed across the entire frontotemporal network, consistent with research that characterized empathic responses as a multi-stage process involving key brain regions including the prefrontal cortex (PFC), insula, and frontoparietal networks.^9,89 The intricate interplay among these neural circuits facilitates the complex cognitive and emotional processes that underpin empathic responses.^{3,15,89,92,93} Recent neuroimaging studies have further specified the roles of specific brain regions in empathic processing, with the anterior insula and anterior cingulate cortex suggested to underlie affective sharing and emotional contagion,^3,15 while the medial prefrontal cortex and temporoparietal junction were suggested to be associated with perspective taking and mentalizing.^6,21

Our focus on the fronto-temporoparietal brain regions revealed enhanced interbrain synchrony in response to empathy-to-distress stimuli compared to neutral stimuli, particularly in areas associated with complex empathic responses in response to empathy-to-distress stimuli. These findings align with the ontogenetic framework of empathy development, where primitive affective resonance gradually matures into sophisticated other-oriented responses.^26,27 The developmental progression toward increasingly complex empathic processing is mirrored in ERP studies, which documented this developmental progression, demonstrating how automatic affective reactions evolve into complex cognitive evaluations through temporally distinct processing stages and reflect different levels of processing.^32,33,34,35 Our results further support the notion of empathy as a distributed process that involves multiple neural systems.⁹⁴ More importantly, our findings extend the understanding of empathy’s neural basis beyond individual brain activity to shared neural representations across individuals. This extension from single-brain to dual-brain dynamics provides empirical support for theoretical models of neural resonance and shared representations in social cognition,^86,88 suggesting that empathic processing involves not only parallel activation of similar brain regions across individuals but also synchronized neural dynamics between them.

We used hyperscanning dual-EEG, a state-of-the-art approach in social neuroscience that enables the simultaneous recording of EEG data from multiple individuals. Hyperscanning data provide a unique perspective on the shared neural processes that underlie complex human capacities, such as empathy, and offer opportunities to address questions beyond the scope of traditional single-brain experiments in social neuroscience.^56,57 We assessed neural alignments based on specifically triggered events of 2,000 ms corresponding to either empathy-to-distress stimuli, analogous to ERP paradigms. This allowed us to evaluate how two physically separated brains synchronized in response to distress across different time windows, ranging from automatic to higher-order processes. Such two-brain perspective allowed us to gain meaningful insights into the neural mechanisms that facilitate shared understanding and joint emotional processes. Our approach represents a significant advance in exploring the neural underpinnings of social cognition and the dynamic nature of empathic processes as they unfold in real time. By adopting this methodology, we contribute to the growing field of interbrain research that is gaining significant traction in neuroscience^95,96 and offer a novel approach to studying the neural basis of social and empathic processing and emotional experiences.

Our investigation focused on interbrain synchrony in the alpha (8–13 Hz) and beta (13.5–29.5 Hz) frequency bands, two neural rhythms strongly implicated in social cognition and emotional processing.^59,97,98 This focus was supported by extensive evidence demonstrating the crucial role of alpha and beta band synchronization across various social contexts. Hyperscanning studies have shown that alpha-band synchrony is enhanced during naturalistic parent-child interactions, modulated by emotional quality and dyadic engagement.⁹⁸ This synchronization is particularly pronounced during early development, where increased alpha coupling between infants and adults occurs during direct gaze and infant-directed speech,⁸² correlating with measures of emotional connection.⁹⁹ In adult interactions, alpha-band coupling has been consistently observed during cooperative tasks¹⁰⁰ and empathic exchanges,⁵⁹ with alpha suppression in frontal, central, and temporal regions marking neural modulation associated with attentional engagement, and processing during empathetic responses to others’ pain.^101,102

Similarly, the beta rhythm has emerged as a critical marker of social connection and empathy. Studies have demonstrated that interbrain beta synchrony underlies empathic communication between both attachment partners and strangers^59,97 in experiencing compassion¹⁰³ and cooperation.^104,105 Beta rhythm has also been found to be involved in active thinking, joint attention, and mentalizing processes triggered by shared coordination.¹⁰⁶ Furthermore, trait empathy and joint attention have been found to predict synchronization levels during face-to-face interactions.¹⁰⁷ In parent-child relationships, beta synchrony was found during both direct encounters and physical separation.^66,67,76 These established patterns of both alpha and beta oscillations in social cognition and emotional processing provided strong theoretical grounds for examining their role in shared neural mechanisms of empathy, even in the absence of co-presence and social cues. Our findings contribute to this growing body of literature by demonstrating how these rhythms facilitate neural synchronization during empathetic responses.

The capacity for empathy, fundamental to human social bonding, and societal formation, emerges from a complex interplay of immediate reactions and sophisticated cognitive processes. Our study makes several novel contributions to understanding the neural basis of empathy. First, we demonstrate the existence of a shared neural mechanism for empathic processing that operates even in the absence of direct social cues, manifesting through interbrain synchronization. Second, we show that while this neural alignment is universal, occurring even between strangers, it is significantly enhanced within attachment relationships, particularly between mothers and adolescents and modulated by biobehavioral factors. Third, our temporal analysis reveals that this shared neural response primarily targets higher-order cognitive components of empathy, specifically occurring in the 1,000–1,500 ms time window, rather than automatic affective responses. This temporal specificity suggests a sophisticated mechanism for shared emotional understanding that goes beyond immediate empathic reactions.

These findings expand our understanding of empathy’s neural architecture by demonstrating how shared emotional experiences can synchronize brain activation between individuals, with relationship status modulating both the strength and temporal characteristics of this synchronization. The enhanced and more sustained neural coupling observed in mother-adolescent pairs provides new insights into how attachment relationships may create complex neural channels for empathic synchrony. Together, these results open new avenues for understanding how interbrain synchrony facilitates complex social processes and supports the development of empathic abilities, while highlighting the fundamental role of close relationships in shaping our capacity for shared emotional experiences.

Limitations of the study

Although our study provides novel insights into the interbrain correlates of empathic response, several methodological considerations warrant discussion. A primary limitation lies in the comparison between mother-adolescent dyads and stranger pairs, where multiple factors were not fully controlled. The groups differed not only in relationship type and familiarity but also in age distribution and power dynamics. Specifically, while mother-adolescent dyads represented hierarchical relationships with significant age and developmental gaps, adult stranger dyads were matched on age and social status. These confounding variables may have independently influenced the observed patterns of neural synchrony beyond the relationship factor alone. Future research should systematically disentangle these variables by examining how age differences, power dynamics, and familiarity independently contribute to interbrain synchrony during empathic processing.

Additionally, while EEG hyperscanning provided valuable insights into the temporal dynamics of shared empathic responses, the methodology’s limited spatial resolution constrains our ability to precisely localize the neural sources of the observed synchronization patterns. Future studies employing complementary neuroimaging techniques, such as fNIRS, could provide more spatially resolved analyses of the brain structures involved in empathic processing. Such multi-modal approaches would help create a more comprehensive understanding of both the temporal and spatial characteristics of interbrain synchronization during empathic responses.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Linoy Schwartz, PhD (Linoy.Schwartz@gmail.com).

Materials availability

The stimuli used in empathy-to-distress and control conditions were compiled from publicly available sources or created for the purposes of this study. Detailed descriptions of the stimuli and the behavioral coding scheme are available from the corresponding author upon reasonable request.

Data and code availability

• •The data generated during this investigation is not publicly accessible due to participant confidentiality considerations; however, it may be obtained from the corresponding author upon submission of a reasonable request.
• •The analytical code utilized in this study is available from the corresponding author upon submission of a reasonable request.
• •Any supplementary materials or additional information pertaining to this research is available from the corresponding author upon submission of a reasonable request.

Acknowledgments

The study was supported by the Simms/Mann Foundation Chair to Ruth Feldman and by the Bezos Family Foundation.

Author contributions

Conceptualization, L.S., J.L., and R.F.; methodology, L.S., J.L, Y.S, and R.F.; investigation and data curation, L.S, J.L, Y.S, C.S, O.H, and O.Z.-S.; formal analysis, L.S and Y.S.; writing —L.S, R.F.; writing—review and editing, R.F, J.L, C.S, and O.H.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, we employed AI tools only to improve the language and readability of the paper, and to generate illustrations for the dyads in separate rooms. After using this tool, we reviewed and edited the content as needed and we take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Software and algorithms
Spyder 5.05
Python 3.8
MNE (v0.17.0)
Autoreject v0.1	Jas et al.108

Experimental model and study participant details

For Study 1, fifty mother-adolescent dyads (N = 100) were recruited through community centers and social media advertisements. Demographic data was available for 47 dyads through self-report questionnaires. All mothers were biological mothers and primary caregivers (Mean age = 43.35 years, SD = 4.04, range = 32.5-56.4). Adolescents (Mean age = 12.03 years, SD = 1.3, range = 9.6-14.6) included 18 males and 32 females, resulting in 18 mother-son dyads and 32 mother-daughter dyads. All participants reported they were mentally and physically healthy, and all children attended state-controlled typical schools.

For Study 2, twenty-five dyads of unacquainted adults (N = 50; Mean age = 27.32 years, SD = 4.62, range = 20-39) were recruited. The sample included 14 males and 36 females. Of the 25 dyads, 13 were same-sex pairs (10 female-female, 3 male-male) and 12 were mixed-sex pairs. All participants reported good physical and mental health. No dyad members had prior contact before the experiment.

In study 1, all mothers signed a written informed consent form for themselves and their adolescent children, and in study 2, each individual signed for themselves. All procedures were explained to the participants prior to the experiment, and the participants were free to leave the experiment at any time with full compensation. Participants were reimbursed for study participation ($30 per hour). The Reichman University institutional ethics committee approved both studies in thus experiment, and all experiments were performed in accordance with the relevant guidelines and regulations.

Method details

Experiment procedure

Participants in Study 1 arrived at the laboratory between 3 PM and 6 PM. This timing was chosen to minimize the effects of changes on oxytocin (OT) reactivity, as research has shown that OT levels differ throughout the day and peak at the afternoon.¹⁰⁹ Upon arrival, participants underwent a 10-minute familiarization period during which no physical contact occurred between mother and child. Following this, baseline saliva samples were collected from both participants. Following saliva collection, the mother and adolescent were separated into two adjacent experiment rooms. This separation was implemented to ensure that no social cues were exchanged between the dyad members during the task.

In Study 2, which did not involve saliva sampling, participants were directed to separate rooms upon arrival. The preparation process, including EEG cap adjustment, took place in these separate rooms. This protocol ensured that the participants had no contact with one another prior to completing the empathy-to-distress task. This separation procedure in both studies was crucial to isolate the neural responses to the empathy task from any immediate social cues or interactions between participants.

Empathy-to-distress task

The study utilized hyperscanning dual-EEG during an 8-minute empathy-do-distress task. During this task, the participants were seated in two adjacent rooms in front of a 24″ S2415H DELL screen with 1920∗1080 resolution and 60-frames-per-second refresh rate. Both computers were wired and connected to the same internal network. The task utilized Psychopy v2021.2.3, to run the task on the computer assigned to the adolescent. The mother’s computer was wired to present the screen of the adolescent throughout the entire experiment, using the VNC viewer 3.18.907 (r38355) x 64 software.

The empathy task comprised eight blocks, evenly divided between empathy-to-distress and control conditions. Each block began with a 10-second contextual sentence presented on screen. For empathy blocks, these sentences described distressing situations (e.g., "The child received poor grades and has now learned that he will have to repeat the school year"), while control blocks presented neutral scenarios (e.g., "The child is reading a book about the history of France and the events that led to the country’s independence". The complete list of contextual sentences is provided in Table S1). Following each contextual sentence, a 1000 ms fixation cross appeared, after which 7-10 stimuli were presented in sequence. The variable number of stimuli per block prevented anticipation of block endings, while maintaining equivalent total stimuli across conditions (Control: 37 stimuli; Empathy-to-distress: 35 stimuli). Each stimulus appeared on the screen for 2000 ms, with a random inter-stimulus interval (ISI) ranging from 1000 to 2000 ms, during which a fixation cross was displayed. This pattern of stimulus presentation and ISI continued throughout the block.

At the conclusion of each block, the adolescent (in Study 1) or Participant 1 (in Study 2) were prompted to answer two questions. The first question was "How easy/hard was it for you to imagine the situation of the people in the images?" Responses were recorded on a 5-point scale ranging from "easy" to "hard." The second question inquired, "How much do you think the people in the images experienced stress or distress?" This was also rated on a 5-point scale, ranging from "a lot" to "a little." At the same time, the mother (in Study 1) or Participant 2 (in Study 2) viewed the other participant’s response from their own computer screen, but did not rate the stimuli themselves.

Dual EEG data acquisition

EEG was recorded simultaneously and continuously for both participants throughout the experiment using a 64-channel BrainAmp amplifier (Brain Products Company, Germany). The EEG system comprised two Brain Product standard subtemporal BrainCaps, each equipped with an integrated chin belt and 32 electrodes. The electrodes were directly attached to the cap and arranged according to the international 10/10 system, an extension of the standard 10/20 system (For a complete list of electrodes and their positions, see Table S2. Theta/Phi coordinates are reported, standardized to a Theta of 90 for the plane through Fpz, T7, T8, Oz).

An analog 0.1–500 Hz band-pass filter has been applied, and data was sampled at 1000 Hz. Electrode impedances were maintained below 10 kOhm, with the ground electrode positioned at AFz. To ensure millisecond-range synchrony between the EEG recordings of the two participants, both EEG helmets were connected to the same amplifier.

Each type of visual stimulus that was presented throughout the task, either Empathy-to-distress or Control, was associated with a specific condition trigger sent to the continuous EEG. This trigger system allowed for offline segmentation and averaging of selected event-related EEG periods for subsequent analysis based on the stimulus type. This setup provided high-quality, synchronized EEG data from both participants, enabling precise analysis of interbrain synchrony during the presentation of Empathy-to-distress and Control stimuli.

Saliva sampling and oxytocin measurement

Participants gave three saliva samples by passive drooling; at baseline upon arrival, immediately after peak training intensity (which occurred following the empathy task) and following a cool-down period. Here, we assessed only the baseline level, which was relevant to the dyad’s state when completing this task. When the participant had difficulty producing sufficient saliva, they were instructed to massage their jaw. All samples were kept chilled and stored at −20°C.

To prepare the sample for measurements, the samples underwent the following procedure:

• 1.Freeze-thaw three cycles: freeze at -80°C and thaw at 4°C to precipitate the mucus.
• 2.Centrifugations at 1500g (4500 rpm) for 30 minutes.
• 3.The supernatant was transferred into a clean tube and stored at -80°C. for at least three days prior the freeze-drying procedure.
• 4.In order to increase the sensitivity, the liquid samples were concentrated four times by freeze-drying for four days to yield a powder. The dry powder is then kept at -20 °C until assayed.
• 5.On the day of assay, samples were thawed completely, re-suspended in sample buffer.

The concentration of OT was determined by Cayman-OT ELISA kit (Cayman Chemicals, Ann Arbor, Michigan, USA). ELISA (enzyme-linked immunosorbent assay) methodology is commonly used for determining hormones concentration in saliva. Samples were measured in duplicates.

Concentration of OT (in pictograms per milliliter) in these samples was determined in duplicates. The assay performed according to the manufacturer’s kit instructions, by using auto-analyzer EVO75 (Tecan Germany). The oxytocin concentration was calculated by Magellan program, according to the kit’s calibration curve. Three internal controls in the low mid and high range were added for quality control. In case of deviation, by more than two SD, the sample from that plate, were excluded and samples were rerun. The inter-assay coefficients of samples and controls were less than 24.1%, which is within the manufacturer reported range.

Quantification and statistical analysis

Data preprocessing

Data preprocessing was conducted using Spyder 5.05 and Python 3.8, utilizing MNE (v0.17.0) for EEG data analysis. Initially, each dyad’s EEG data file was split into two separate files, one for each participant, to allow for individual preprocessing. All data files then underwent average-referencing, and a 1-50 Hz bandpass filter has been applied. In line with previous studies.^59,110 Subsequently, the EEG data were segmented into time-locked events based on stimulus appearance triggers (either Empathy-to-distress or Control stimuli), creating 2000 ms long event-related epochs, baseline-corrected to 200 ms prior to stimulus onset. To remove artifacts, we employed Autoreject v0.1,¹⁰⁸ an unsupervised algorithm using Bayesian optimization, which eliminated trials containing transient jumps in isolated EEG channels and artifacts affecting channel groups. Further artifact removal has been accomplished using Independent Component Analysis (ICA). We used MNE’s implementations of fastica and CORRMAP¹¹¹ to remove systematic physiological artifacts. Independent components (ICs) related to eye movements and muscular artifacts were manually selected for exclusion and served as templates for identifying and removing similar components across all participants. This process targeted eye blinks, eye movements, and muscle artifacts (see Figure S1 for visualization of excluded IC components).

In study 2, the data of 2 dyads was not analyzed due to technical issues during data acquisition (i.e., electrode malefactions and connectivity issues), and data of one dyad was excluded from analysis due to containing voltage fluctuations of > 100 μV in channels of interest, resulting in the analysis of 22 dyads.

Interbrain synchrony calculation

We quantified interbrain synchrony using the weighted phase lag index (wPLI), a connectivity method previously employed in a large body of studies of naturalistic social interactions.^{60,66,67,76,84,110} Our analysis focused on two rhythms – the alpha rhythm (8-13 Hz) and the beta rhythm (13.5-29.5 Hz). Analytic signals were computed using IIR filtering with a Hamming window to mitigate distortion and border effects, followed by the Hilbert transform.¹¹² In line with prior research,^{59,62,66,67,76} we divided the EEG cap into pre-defined regions of interest (ROIs) based on our research hypotheses, concentrating on frontal and temporo-parietal areas. We established four ROIs, each comprising three electrodes: right frontal (RF: Fp2, F4, F8), left frontal (LF: Fp1, F3, F7), right temporal (RT: T8, TP10, P8), and left temporal (LT: T7, TP9, P7). This grouping of channels into regions enhances the reliability of region specification and provides a more meaningful interpretation of the results.

For analysis of interbrain synchrony, we assessed the combinations of the 4 ROIs (RF, LF, RT, LT), resulting in 16 possible combinations of linkages between the ROIs across participants. We calculated the wPLI value for each dyad’s ROI pair as the mean connectivity value of the nine possible electrode combinations (3 electrodes in one ROI × 3 electrodes in the other ROI). Finally, to determine the interbrain connectivity of the entire fronto-temporal network, we averaged all 16 possible linkages into a single value. This measure represented the overall interbrain connectivity across the shared network for each stimulus type (Empathy-to-distress, Control). This approach allowed for a comprehensive assessment of the dyads’ interbrain synchrony across key regions implicated in empathic processing in our main analyses, while also providing a robust measure of overall network connectivity between participants while viewing the different stimulus types.

Behavioral coding

To assess brain-behavior correlations, the mother and adolescent behaviors during a following face-to-face interaction,⁶⁶ has been coded offline using the Coding Interactive Behavior manual scheme (CIB).¹¹³ The CIB is a well-validated rating system used for coding social interactions that has yielded over 200 publications across multiple cultures, age range, and pathological conditions,^72,81,114 including hyperscanning research.^{59,66,76,84,110}

The behavioral coding of the dyads was conducted offline, by trained coders who were blind to study hypotheses with inter-rater reliability for 20% of the interactions exceeding 90% on all codes (intra-class r = .93, range = .89-99).

The CIB yields overall a total of 52 codes, each rated on 5-point scales that aggregate into theoretically-based constructs. Here, we used the CIB construct of Dyadic Behavioral Synchrony, which comprises the following CIB scales: Dyadic reciprocity, Dyadic flow, and Dyadic relaxed state (reverse coded from dyadic tension). The dyadic behavioral synchrony construct was found to be linked with dyadic interbrain synchrony in previous research.⁵⁹

Statistical analysis

Statistical design overview

Statistical analyses were performed using repeated-measures analysis of variance (ANOVA) with within-subject factors of Stimulus Type (Empathy-to-distress vs. Control) and Time Window (0-500, 500-1000, 1000-1500, and 1500-2000 ms post-stimulus). For spatial specificity analysis, ROI Combination (16 levels) was included as an additional within-subject factor. Study comparisons utilized mixed ANOVA with Sample (Mother-adolescent vs. Stranger dyads) as a between-subjects factor.

Comparing neural synchrony between empathy-to-distress compared to control stimuli

Building on the rich body of previous ERP studies, which evaluated the time-sensitive temporal dynamics of empathic responses, we assessed the dyads' interbrain synchrony in four different time windows: 0-500 ms (automatic responses and early cognitive processes), 500-1000 ms (intermediate cognitive processes, involving emotion regulation), 1000-1500 ms (complex higher-order cognitive assessments), and 1500-2000 ms (late neural responses involving high-order cognitive processing).

To assess whether processing Empathy-to-distress stimuli induced greater interbrain synchrony compared to neutral control stimuli, we conducted our primary analysis on the neural connectivity values (wPLI) representing the shared interbrain synchrony of the frontotemporal network of the dyads. We employed a repeated-measures analysis of variance (ANOVA) on these wPLI connectivity values. The ANOVA incorporated two within-subject factors: Stimulus Type (Empathy-to-distress stimuli, Control stimuli) and Time window (0-500 ms, 500-1000 ms, 1000-1500 ms, 1500-2000 ms following stimulus onset). This approach allowed us to examine the effect of the type of stimulus on dyadic interbrain synchrony across the entire 2-second presentation period, divided into four 500-ms intervals, reflecting different processes of empathic responses, ranging from automatic to complex cognitive processing.

Following our primary analysis of the overall fronto-temporal network interbrain connectivity, when a significant main effect for Stimulus type (Empathy-to-distress vs. Control) was observed, we conducted a detailed examination of the 16 individual interbrain connections between predefined regions of interest (ROIs). Using repeated-measures ANOVA, we analyzed these connections across the two stimulus types, allowing us to examine differential connectivity patterns.

Published: May 12, 2025