“Resonance” is a commonly-used term that describes the human experience of powerful, connecting and activating interactions (Duarte, 2013). For instance, we can “resonate” with a film or with a new friend. Metaphors related to resonance are also common, such as in the expressions “syncing up,” “getting on the same wavelength,” or even “feeling good vibes.”

Although the term “resonance” is often intended as a metaphor to describe an interaction, in many cases physical resonance may also be a mechanism underlying the interaction. For instance, people metaphorically “resonate with music” but the brain also physically resonates with music: the actual frequencies of sound and rhythm can be observed in the frequencies of electrical activity in the brain (Coffey et al., 2019, 2021; Kaneshiro et al., 2021; Pandey et al., 2021). Or, consider the common expression “syncing up” to describe a meeting. Even though “syncing” (that is, synchronization) is intended purely as a figure of speech, human communication does link to measurable “inter-brain synchrony” (Dumas et al., 2010; Dikker et al., 2019; Czeszumski et al., 2020; Kingsbury and Hong, 2020; Dumas and Fairhurst, 2021; Moreau and Dumas, 2021). This article aims to create a bridge between the human experience of resonance and resonance as a physical mechanism. In so far as resonance is more than a metaphor—if resonance is also a causal mechanism in human interactions—then this will have implications for measurement and design.

Some skepticism is justified in viewing resonance in human interactions as “real” rather than as just a metaphor. Historically, sympathetic resonance (sympathy: σv^μπα^θεια^ or sumpátheia) was viewed as the primary mechanism for magical phenomena. For instance, the neoplatonic philosopher Plotinus (205–270) wrote: “But how do magic spells work? By sympathy [sumpatheiai] and by the fact that there is a natural concord [sumphonian] of things that are alike [homoion] and opposition of things that are different.” (Lobis, 2015). Even in the modern era, there remains a widespread belief system that positive thinking or “thought vibrations” can bring about positive real-world occurrences through sympathetic resonance (Atkinson, 1906; Hicks and Hicks, 2006; Ehrenreich, 2009). Perhaps as a result of this association with magic, resonance was not always acceptable as a scientific explanation. A recent column in Nature Physics notes:

Now, the situation has changed: the term “resonance” is abundant in contemporary scientific literature (reviewed in Box 1). However, the term is often used ambiguously, where it is unclear whether “resonance” is being treated as a metaphor or as a physical mechanism. This ambiguity is present in the social sciences as well as in physics. Resonance in physics is an increasingly broad concept that refers to a range of phenomena. To bring clarity, we offer a glossary in Box 2 with proposed definitions for resonance and related terms, such as synchronization, entrainment, reverberation, etc.

To provide grounding for resonance in human dynamics, this section outlines physical resonance as a causal mechanism in acoustics. Though we focus on sound, it is important to note that resonance operates in all oscillating systems, regardless of medium. This universality results from the fact that resonance is a mathematical property—it is the natural result of the alignment of phases in oscillating systems.

A wine glass offers an excellent example of the physics of resonance. First, if a glass is gently tapped with a spoon, there will be a reverberating sound that reflects the natural frequencies of oscillations in the wine glass. These natural frequencies, which are inherent to the structure of the glass, are also known as characteristic frequencies or eigen frequencies (“eigen” is German for “own” or “inherent”). These natural frequencies are also the resonance frequencies of the glass: when external, forced oscillations match these natural frequencies, resonance occurs. But, while tapping the glass with a spoon may reveal the resonant frequencies of the glass, the wine glass is not in resonance with the spoon.

Sympathetic resonance occurs when external, forced oscillations are aligned to a system's own natural oscillations. If a loudspeaker plays the resonant frequencies of a wine glass, the glass will begin to oscillate at much greater amplitudes than if the speaker played other, non-resonant frequencies. Now the glass is in resonance with the speaker. This effect manifests in other common acoustic systems, as well. When one tuning fork is struck near another identical fork, they will both begin to oscillate together, having been coupled together in synchrony via the acoustic vibrations. Similarly, two strings tuned to the same note will move one another in synchrony through sympathetic resonance (Figure 1).

The relationship between resonance, synchrony and amplification was articulated by German nineteenth century scientist Hermann Helmholtz. His book “On the Sensations of Tone as a Physiological Basis for the Theory of Music” (Helmholtz, 2009; originally published 1863) offers the first scientific exposition of sympathetic resonance in acoustics. His primary illustration of sympathetic resonance involves the resonance between a church bell and its bellringer. If the bellringer provides consistent pulls at a frequency that aligns to the bell's natural rate of swinging, then the swinging will be rapidly amplified. Importantly, the sympathetic resonance occurs when phases of oscillation align: that is, when the downward pull of the bellringer matches up with the downward motion of the bell's swing. The role of synchrony in sympathetic resonance is easier to observe with a slow bell ringer than with the rapid oscillations of a wine glass. Yet, even the sympathetic, synchronized oscillations of a wine glass can be made visible with high frequency camera equipment (Slow Mo Guys, 2021).

Synchrony between systems does not necessarily imply sympathetic resonance. Two systems might be synchronized with each other due to a third system, for instance, or for other non-causal reasons (Hasson and Frith, 2016). Other forms of resonance only occur with powered oscillators (like the clocks and metronomes in Figure 1), namely entrainment and synchronization. These terms—which explain phenomena like the synchronization of fireflies or the entrainment of dancers to a musical beat— are defined and discussed in Box 3. In this article, we treat these two terms as subsets of sympathetic resonance (by analogy, like squares are subsets of rectangles).

Having briefly considered the operation of resonance as a physical mechanism, we now wish to bring clarity to the metaphorical use of resonance in science and broader culture.

A recent review of the word “resonance” in the language of scientific literature (Ruthven, 2021) reveals that resonance typically serves as an implicit metaphor to indicate 1. agreement (e.g., new evidence can resonate with an existing theory), 2. arousal (e.g., a film that resonates is engaging and moving) or 3. action (e.g., the resonance of a speech can motivate people to take action). But, despite a vast number of scientific articles that use resonance as a term, it is only very rarely defined. The lack of definition suggests that “resonance,” as a term, is easily and broadly understood intuitively as a metaphor.

Metaphors are useful when they enable concrete, familiar experiences to communicate abstract, conceptual meanings (Lakoff and Johnson, 1980; Yang, 2014). A metaphor involves the pairing or alignment of concepts between a concrete source and a more abstract target; this coupling of concepts produces mappings that allow multiple concepts to be integrated together into an emergent space of meaning (Holyoak and Stamenković, 2018). An example set of metaphorical mappings use the metaphor “Love is a Journey”: for instance, in a journey (the source) there are travelers while in love (the target) there are lovers. In Table 1, we provide a similar set of explicit mappings between the metaphor of acoustic resonance and resonant human interactions.

With this introduction to resonance established, we are now posed to explore the alignment between the metaphorical experience of resonance (as in a film that resonates) and the physical phenomena of resonance itself. The next section of the article considers the research basis for understanding resonance in human interactions.

The previous example shows how rhythmic human activity can be investigated as a context for physical resonance phenomena. Rhythmic human interactions clearly occur in artistic domains, such as music-making (Clayton et al., 2005), dance (Larsson et al., 2019) and various kinds of cultural rituals like chanting (Gelfand et al., 2020). Rhythmic interactions are also common in everyday life, as in the case of walking, conversational turn-taking (McGarva and Warner, 2003; Wilson and Wilson, 2005; Lee et al., 2010), patterns of eye contact (Wohltjen and Wheatley, 2021) or with interactions like handshakes (Melnyk and Hénaff, 2019). More intimate rhythmic interactions occur during human sexual behavior (Safron, 2016). Leading up to the moment of birth, midwives often advise expectant mothers to push in phase synchrony with their own rhythmic uterine contractions (Hanson, 2009). Researchers have also observed that the earliest interactions between parent and child are strongly rhythmic (Stern et al., 1985). Babies cry in a rhythmic manner and caregivers soothe them with synchronized motions (Trehub and Trainor, 1998). Synchrony in caregiving appears to literally “tune” the human social brain (Yaniv et al., 2021).

What makes human behavior so rhythmic? Human rhythms are believed to emerge from a broad range of biological oscillators that are present across the brain and the rest of the body (Varga and Heck, 2017). For instance, rhythmic central pattern generators in the spinal cord not only drive locomotion (Ijspeert, 2008; Guertin, 2013), but also drive heart rhythms (Bucher et al., 2015) and breath rhythms (Molkov et al., 2014). Furthermore, body and brain oscillations appear to integrate together in a hierarchical architecture (Klimesch, 2018).

Rhythmic—and resonant—phenomena are found in the brain at multiple levels, from neurons to circuits to brain waves (Buzsaki, 2006). Individual neurons have natural intrinsic oscillatory periods (Lampl and Yarom, 1997); neurons can be tuned to respond to different input frequencies through resonance, like “strings on a violin” (Das et al., 2017). At the level of neural circuits, the reverberation of recurrent activity in neural ensembles plays a key role in memory (Wang, 2001; Tegnér et al., 2002; Han et al., 2008), as originally predicted by neuroscientist D.O. Hebb in 1949 (Hebb, 2005). Finally, large-scale electrical oscillations in the brain, or brainwaves, demonstrate clear resonance effects (Herrmann, 2001) that are observable through electroencephalography (EEG). Further to this, neurobiological processes associated with adaptive learning (Grossberg, 2013, 2017), perceptual learning (Raja, 2020), and ecological cognitive architecture more generally (Raja, 2018), have all been theorized as forms of physical resonance.

Neurons are, technically speaking, non-linear oscillators (Izhikevich, 2007; Stiefel and Ermentrout, 2016)—therefore, it is not so surprising that large collections of neurons exhibit both internal resonance effects (one part of the brain resonating to another) and external resonance effects (the brain resonating to environmental phenomena). Many scientists believe that physical resonance in the brain plays a major role in music perception, such as the physicist and neuroscientist Ed Large, who claims:

Some researchers have recently proposed a unified account of rhythmic synchronization and entrainment in the brain (Lakatos et al., 2019); other researchers have proposed a unified account of the biological, neurological and physical mechanisms involved in the “rhythmic entrainment of biological systems” (Damm et al., 2020). Rhythmic entrainment has been found to govern patterns of interaction at a social, population, and even species level—where, in the latter case, the entrainment of natural oscillations can be observed at the scale of economies and ecosystems (Greenfield et al., 2021). In short, it would appear that resonance effects can operate all the way up and all the way down: from neurons to economies.

Human interactions can naturally synchronize through the process of entrainment (Boxes 2, 3), which is akin to the natural synchronization of metronomes (Figure 1). Social neuroscientist Ruth Feldman (2012, 2017) argues that biobehavioral synchrony (in behavior, heart rate, endocrine production and brainwaves) serves as a key principle underlying parental love, romantic love, friendship and human attachments. Indeed, when loving human partners interact, their rhythmic communication produces measurable physical synchronization in behavior (Grafsgaard et al., 2018), in heart rate (Prochazkova et al., 2022) and in the brain (Kinreich et al., 2017).

Some argue that the ability to synchronize to a beat is one of the core skills associated with human social behavior. Kirschner and Tomasello (2009) found that children 2–4 years old could adjust their natural drumming tempo to match another beat—but that their accuracy in synchronizing was significantly higher when they drummed with a human partner (as opposed to drumming along with a machine or drumming along with a drum sound produced by a speaker). The authors argue that “drumming together with a social partner creates a shared representation of the joint action task and/or elicits a specific human motivation to synchronize movements during joint rhythmic activity.”

Humans are typically much less able to synchronize to rhythms of visual flashes than to rhythms of auditory tones (see review by Repp and Su, 2013). But, rhythmic entrainment and synchronization is not specific to music or auditory experiences. With certain forms of visual stimuli (i.e., bouncing balls), visual synchronization becomes nearly as accurate as auditory synchronization (Iversen et al., 2015). Researchers have also found that deaf individuals exhibit enhanced synchronization to visual rhythms, suggesting that the ability to attune to rhythms is at least partially based on experience and not just a result of biological coupling between the auditory and motor system (Iversen et al., 2015). Researchers have found that humans can synchronize to tactile pulses on their back with higher accuracy when feeling the vibrations played over their entire back rather than at just a small portion; similarly, rhythms that engage multiple sensory modalities also produce more accurate synchronization (Ammirante et al., 2016). This suggests that overall sensory immersion and attentional engagement affects the propensity to synchronize with rhythms.

Synchronization helps support coordinated actions between individuals. Meta analyses (Morgan et al., 2017) have shown that behavioral synchrony in groups increases prosocial behavior, increases perceived social bonding, and generally feels good (as measured as increases in positive affect). Why might behavioral synchrony feel good? Cracco et al. (2021) claim that “synchrony is aesthetically pleasing and a signal of group cohesion, as stimuli that are processed more fluently are known to produce a hedonic response.”

Rhythmic synchronization is very rare in the animal kingdom, at a social level, apart from special examples (see extended discussion in Box 4). The human capacity for rhythmic synchronization may have coevolved in human cultures because it enhanced social bonding between sexual partners, between parents and children, and within larger social groups. Savage et al. (2021) state:

The scientific understanding of rhythmic entrainment and neural resonance is a fast-moving area of neuroscience that is being propelled by new hyperscanning methods that scan the brains of multiple interacting participants simultaneously. Interpersonal neural synchrony at the group and dyadic level has been shown to be associated with a number of predictors, including shared stimulus features, joint actions, personality traits, social intentionality, relationship quality, and cooperation (see, e.g., Czeszumski et al., 2020 for review).

For example, recent work from co-authors of this article investigated the relationship between inter-brain synchrony and group dynamics and found that EEG inter-brain synchrony predicted collective performance among teams better than self-report (Reinero et al., 2021). In another line of work, group-based inter-brain coherence predicted class engagement and social dynamics in groups of high school students during their real-world lessons (Dikker et al., 2017; Bevilacqua et al., 2019). Social closeness with the teacher also correlated with brain-to-brain synchrony—that is, enhanced synchrony was found with students who reported greater engagement with the teacher. Finally, and perhaps most directly related to the concept of resonance: Brainwaves of students who engaged in face-to-face interactions before class were more synchronized during class, even if students were no longer interacting. This finding raises interesting questions about the role of resonance in the directionality of the relationship between human face-to-face interaction, inter-brain synchrony, and social connectedness.

Humans may be predisposed to synchronize with each other, but this does not always lead to positive outcomes. Synchronization also has some important tradeoffs; Gelfand et al. (2020) claim that synchrony can produce conformity, destructive obedience, groupthink, antisocial aggression and also impair group creativity. They point to findings (Wiltermuth, 2012a,b) that people who have been randomly assigned to a synchronous activity are more likely to comply with an anti-social order (e.g., irritating a stranger) and to follow a morally compromised command (in the study, participants were asked to grind up live bugs). Synchrony also increases the likelihood that people will engage in conformity, like copying majority opinions rather than following their personal preferences (Dong et al., 2015). Further, sometimes synchrony is simply “situationally inappropriate;” in a study of a complex verbal coordination, groups that were randomly assigned to a synchronization task performed worse, reported higher levels of conflict and reduced group cohesion (Wood et al., 2018).

Gelfand et al. (2020) randomly assigned participants to march synchronously around a college campus or at their own pace. The participants who synchronized showed reduced creativity when writing stories. They also found that synchronous marching discouraged the development and sharing of minority perspectives during decision-making. They relate this finding about synchrony to the need to balance “tightness” and “looseness” in culture.

The ability to flexibly move in and out of synchrony appears to be critical to adaptive flexibility. Mayo and Gordon (2020) claim that “two tendencies exist simultaneously, one to synchronize with others and another to move out of synchrony and act independently. We suggest that an adaptive interpersonal system is a flexible one, able to continuously adjust itself to the social context.”

Savage et al. (2021) point out the key difference between rhythmic integration and pure synchronization: rhythm is predictable but also flexible to accommodate diverse individual contributions. This is because rhythm involves two essential components: 1. equally timed beats (isochronicity) and 2. a hierarchical structure (meter). “While synchronization solely to the beat (e.g., in marching or unison chanting) allows large groups to integrate, it tends to submerge individual contributions. Meter solves this problem by allowing many individuals to contribute, out of phase, to the same integrated rhythm.” Social rhythms (of speech, music, dance, etc.) can thus support diversely coordinated actions within a loosely unified structure.

Sympathetic resonance—including synchronization and rhythmic entrainment—appears to have been a key factor in human evolution (Savage et al., 2021; Lin and Lomas, 2022). Resonance relates in a fundamental way to the human capacity to feel what another person feels, which is often called empathy. But, before the term “empathy” was coined in the twentieth century, the ability to feel what others feel was referred to as “sympathy”—as in sympathetic resonance. The eighteenth century philosopher Adam Smith wrote his first book, “The Theory of Moral Sentiments” (Smith, 1759), with the general thesis that “sympathy” accounts for a large portion of moral behavior. Specifically, he explained that people like to help other people because they sympathetically feel good when other people feel good and sympathetically feel bad when other people feel bad (Schliesser, 2015).

Later, the nineteenth century German psychologist Theodore Lipps used the German term Einfühlung to describe how people “feel into” the states of other people and even art pieces. By using an inner imitation or simulation, people seem to be able to fuse with artworks or persons through a process of “Psychische Resonanz” (Lipps, 1891). For instance, watching a tightrope walker produces a resonance with internal associated feelings like vertigo. The representation of the performer in one's own mind allows one to feel how oneself would feel in the same situation. The psychologist Edward Titchener reviewed Lipps' work in 1909 (Titchener, 1909) and, rather than using the German Einfühlung, he coined the new English word Empathy (Schliesser, 2015).

Empathy is viewed as a critical component of human social interactions. However, it is extremely challenging to pin down. While there is an enormous amount of scientific work on empathy, there is still considerable debate about its definition (Hall and Schwartz, 2019). Is empathy a singular capability or does it result from a “laundry list” of characteristics? Psychologists generally accept the division between cognitive empathy and affective empathy. Cognitive empathy refers to the ability to recognize and understand another person's mental state (cognitive processes captured by what is referred to as “theory of mind” or mentalizing), while affective empathy refers “the ability to vicariously experience the emotional experience of others” (Reniers et al., 2011). Furthermore, psychologists will often draw another distinction between empathy (which involves the ability to distinguish the experience of another person's emotion from one's own emotional state) and emotional contagion. Emotional contagion involves the direct propagation of emotional states; unlike empathy, this effect is common in non-humans, like mice (Hernandez-Lallement et al., 2020).

Regardless of definition, the capacity for empathy (or, at least, affective empathy) is typically conceptualized as emerging from motor resonance. Motor resonance describes how the spatial-temporal activations of an observer's brain mirrors the brain activations of another person as they perform some set of actions. That is, when observing the physical behavior of another person, the brain regions related to this behavior activate in both the observer and the person enacting the behavior, creating a sort of spatial-temporal synchrony between observers and actors. Thus, motor resonance is a type of physical resonance that provides a mechanism for sharing conscious experiences between people. For instance:

Several theories of empathy describe motor resonance as the mechanism underpinning the mirroring processes of emotions and actions, where mirroring processes are defined as automatic processes for internal imitation (Iacoboni, 2009) or embodied simulation (Gallese, 2009). The simulation theory of empathy (e.g., Preston and De Waal, 2002) suggests that people can feel what other people are feeling because observing another person's behaviors will coactivate or call upon neural representations of one's own bodily experience (Hurley, 2008).

While stereotypical expressions may produce meaning symbolically (e.g., a smile is symbolically associated with joy or a frown with sadness), human emotional expression is far more dynamic, expressive, and context dependent. The spatial-temporal dynamics of, say, a rapidly lifted eyebrow can express inner emotional states with great specificity. Observing the eyebrow rapidly lifting will engage our own motor cortex to activate in a similar time scale and specifically in the areas expected for the muscles involved. These spatial-temporal activations appear capable of automatically triggering associated emotional states (Wood et al., 2016). That is, whatever feelings might have been associated with rapid eyebrow lifting in the past, either in the self or in others, will now be primed. In this manner, interpersonal motor resonance appears to support “mind reading” (Agnew et al., 2007) and the sharing of conscious experiences (Lin and Lomas, 2022). Because our brains reflect or mirror each other, through resonance, we can sympathetically experience the feelings associated with other's actions, in part by knowing “how we would feel if we were acting that way.” And it is not just observation: even listening to descriptions of actions can trigger motor resonance (Zwaan and Taylor, 2006).

Researchers continue to debate the origins of mirroring processes, but they appear to result from simple bidirectional associations between perception and motor responses that are learned over time (Keysers and Gazzola, 2009, 2014; Hanuschkin et al., 2013). Simple correlations of associated actions and observations seem to produce “action perception circuits” that serve as the neural mechanism for mirroring processes (Pulvermüller, 2018).

For a clear example, fMRI results show that when people watch others perform actions with their hands, mouths or feet, there are activations in their own premotor cortex—activations that are also triggered when performing those actions themselves. Furthermore, these action activations occur “following a somatotopic pattern which resembles the classical motor cortex homunculus.” (Buccino et al., 2004a) For instance, if person A watches person B kick a ball, the “leg part” of the premotor cortex will show a similar pattern of activation in person B (the kicker) and in person A (the observer).

If we consider the sympathetic resonance of two tuned strings, the strings have in common their natural frequencies of oscillation. A similar sympathetic resonance occurs when we see another person smile; this can trigger similar action representations in our brain and activate associated emotional states. Following the metaphor of two tuned strings, the “like me” hypothesis predicts that the degree of motor resonance between an actor and observer will correlate with the degree of similarity between the actor and the observer.

Buccino et al. (2004b) investigated human motor resonance in response to dogs, monkeys and people. They found that “Actions belonging to the motor repertoire of the observer (e.g., biting and speech reading) are mapped on the observer's motor system. Actions that do not belong to this repertoire (e.g., barking) are essentially recognized based on their visual properties.” In other words, actions that are not “like me” may be recognized but they do not resonate.

If resonance is enhanced when observing actors similar to the observer, is it also impaired when there is a lack of similarity? Researchers have found that when subjects observe people of a different ethnicity, there is significantly less motor resonance than when watching members of the same ethnicity (Gutsell and Inzlicht, 2010; Azevedo et al., 2013). This unfortunate effect is predicted by resonance theory: less similarity, less resonance.

The resonance between actions is dependent upon a person's ability to do those actions. The “like me” hypothesis predicts that persons who are highly trained in a particular skill should be able to resonate with another person trained in the skill, at least to the extent that their action-observation circuits are mutually developed. Work with expert dancers using fMRI (Calvo-Merino et al., 2005; Cross et al., 2006), EEG (Orgs et al., 2008), and facial EMG (Kirsch et al., 2016a) provides evidence in support of this idea that shared learning experiences and shared skills will increase motor resonance.

Evidence against the Like-Me hypothesis comes from the fact that similarity does not always enhance activation intensity. A series of fMRI experiments showed that mirroring processes (also known as the “Action Observation Network” or AON) are more strongly engaged during the observation of robot-like motions, both when the motions were performed by actual robots and when people act in a jerky, robot-like manner (Cross et al., 2012). It appears that the relationship between familiarity and neural resonance is not entirely linear because—in part—the perception of novelty (Knight and Nakada, 1998) also amplifies the brain's response to actions (Gardner et al., 2017).

The brain's sensitivity to novelty may help explain why motor resonance is exceptionally amplified when expert dancers observe other expert dancers perform (Cross and Ramsey, 2021). Experts not only have deep familiarity with the movements but they will also have an expert sensitivity to the many small novelties within the expert's individual execution. Familiarity and novelty—though seemingly opposite—both contribute strongly to an aesthetic experience (Hekkert et al., 2003). This may account for why aesthetically valued actions influence motor resonance. Researchers have found that the intensity of activations in the AON (a brain marker of motor resonance) correlates with aesthetic ratings of the observed dances (Cross et al., 2011). Calvo-Merin et al. (2008) also examined the neural response to dance movements and noted that, of five aesthetic dimensions (like-dislike, simple-complex, dull-interesting, tense-relaxed and weak-powerful), only shifts in liking-disliking correlated with the brain response in the AON.

Thus, several factors can enhance motor resonance, including inter-subject similarity. Generally speaking, motor resonance appears to be enhanced by the overall motivational relevance of the subject-observer interaction; the similarity of the subject, the novelty of the interaction, the aesthetic quality of the interaction, and when the subject is viewed as desirable, powerful or sharing a common goal (Greenberg, 2019). This also implies some possibilities for breakdown and pathologies in psychological relationships due to misattunement in interpersonal resonance (Bolis et al., 2017).

To conclude this section on the role of resonance in human interactions, consider this quote by the actor Will Smith explaining his growth as an actor during his performance as Richard Williams, the father of Venus and Serena Williams. The quote illustrates the connection between the body, the communication of emotional depth and how resonance is generated during aesthetic experiences (described here as “vibrations”). It also shows the vast distance roboticists must travel to approach the capabilities of human actors in producing effective sympathetic resonance (for resonance in acting, also see Bogart, 2021).

Dance and music have been an important driver of social robotics. Kozima et al. (2009) used a yellow dancing robot “KeepOn” to either dance in synchrony with background music or dance out of synchrony. They found that, when dancing in synchrony, children were more likely to socially interact and to do so for longer. While there are many dancing robots (reviewed in Bi et al., 2018), most involve pre-programmed motions that are unable to adjust to external visual or auditory stimuli. Responding in real-time to external motions (such as a human dancer) is often limited by hardware and software processing delays. Behavioral resonance is computationally challenging.

Nico, a drumming robot (Crick et al., 2006), used visual, auditory and proprioceptive data to “attune to a tempo that is set by a human conductor, in concert with human performers.” To accomplish this, Nico uses multiple oscillators that model a hierarchy of rhythmic attention. To detect the beat, Nico used cameras to follow the ictus of the conductor's hand (which is when it “bounces” off an imaginary line). As hardware constraints prevented Nico from following a faster tempo, faster beats caused the robot to find a lower hierarchical level of the rhythm: a tempo half that of the beat. The researchers found that human musicians playing along learned to accommodate Nico's mistakes and attune to them. More recently, another successful synchronizing drumming robot was demonstrated by Iqbal and Riek (2021). Though these systems involved simple rhythmic beats, “Shimon” is a robotic marimba player that plays along with human accompaniment in a variety of ways, including call and response (Hoffman and Weinberg, 2010). Motivated by the notion of robotic movement as a dynamic affordance (Hoffman and Ju, 2014), Shimon has an expressive non-humanoid head that synchronizes with the musical beat. Arguably, Shimon vibes with other human players.

Beyond affective interactions, entraining to rhythms can support basic robotic locomotion. Walking on two feet is extremely difficult for robots unless it is on a flat predictable ground (Endo et al., 2008). Some roboticists have found success in using a biomimicry approach—they use “central pattern generators,” like those in the human spinal cord, to achieve dynamic stability through rhythmic entrainment (see reviews, Ijspeert, 2008; Buschmann et al., 2015; Aoi et al., 2017; Xie et al., 2021). This oscillatory approach to robotics also applies to robotic prosthetic limbs for humans, where synchrony with the human motion must be extremely precise (Ronsse et al., 2010).

Synchronization and entrainment have also been useful for the implementation of deceptively simple motor movements, like shaking hands (Melnyk and Hénaff, 2019) or handing a ball to another person. Using the robot iCub, Duarte et al. (2021) used a coupled dynamical system to learn the motor resonances between arm motions during the handover of the ball. Ansermin et al. (2016) also used a coupled oscillator approach to enable the robot NAO to imitate human gestures through entrainment and synchronization. The researchers found that mutual entrainment between the robot and human enabled gesture mirroring and precise synchronization with far fewer computational resources than other approaches (e.g., those involving a high-level planning process).

Robot imitation is useful for many reasons, whether for helping robots learn through demonstration (Argall et al., 2009) or to make robots more persuasive (Bailenson and Yee, 2005). Robots can imitate humans in many ways—but usually in ways that are very different from how humans imitate each other (Breazeal and Scassellati, 2002). Robots that use oscillators to resonate or synchronize with people is a more limited approach but often useful. Using a “mirror neuron framework,” Barakova and Lourens (2009) gave simulated as well as embodied robots the ability to synchronize with human movements; this led to improved turn taking behaviors. Kopp (2010) used a motor resonance approach to support intentional alignment between robots and people. Researchers have proposed a variety of methods for the quantitative measurement of synchrony in human interactions (Delaherche et al., 2012) and the measurement of motor resonance between humans and robots (Sciutti et al., 2012). These approaches have been useful for demonstrating the presence of motor contagion between people and robots (Bisio et al., 2014). A motor resonance system successfully enabled a robot to learn from a human demonstrator to introduce itself using Taiwanese Sign Language (Lo and Huang, 2016). Coupled oscillators, based on central pattern generators, enabled the robot Pepper to wave back at a human partner in an adaptive, synchronized manner; this was perceived as more enjoyable than a non-adaptive wave (Jouaiti and Henaff, 2018).

However, not every application of movement synchrony enhances outcomes. For instance, Henschel and Cross (2020) conducted a controlled experiment to investigate how synchronized task behavior affected the likeability of the humanoid robot Pepper. They found that synchronized task performance had no effect on the likeability of the robot. This was surprising in light of contemporary attitudes:

Robots can influence the biorhythms of people; for instance, Macik et al. (2017) showed that a non-humanoid robot can help entrain breathing patterns while Sato and Moriya (2019) used AI-controlled music tempo to control changes in the rate of breathing. Robots that promote sleep using rhythmic entrainment include the Somnox Sleep Robot (Mohammadi-Khanaman and Lundström, 2019) or the Fisher-Price “Soothe'n'Snuggle” stuffed toy (Figure 3), which uses rhythmically pulsing movements, sounds and lights to help small children fall asleep. While the “Lulladoll,| which plays breathing and heart-beat sounds, did not have a significantly beneficial effect on infant sleep (O'Loughlin, 2018), the Philips Smart Sleep system did produce improvements in slow wave sleep and executive functioning in adults (Diep et al., 2020). This headband system uses EEG and audio-pulses to create a closed loop system that entrains slow waves associated with deep sleep.

In certain situations, people appear to automatically align their speech and behavior to artificial partners. This synchronization has been shown through alignments in speaking rate (Bell et al., 2003), prosody (Suzuki and Katagiri, 2007), gestures (Iio et al., 2011), gestural rhythm (Ansermin et al., 2017), formality of speech (Kühne et al., 2013), vocabulary (Iio et al., 2015) and facial expressions (Hofree et al., 2014). When this alignment occurs, it tends to be associated with a positive experience of the artificial partner. For instance, Fujiwara et al. (2021) showed that when humans spontaneously synchronized their motions to a non-human partner, humans were more altruistic and reported greater affiliation for their non-human partner. Importantly, synchronization effects are highly dependent upon the specific social context—for instance, a competitive task can easily produce a reversal in facial expression synchrony, like a winner smiling at a losing frown (Hofree et al., 2018).

Robots can also serve as a medium or platform to help synchronize people together. For instance, the BAO-ME is “a zoomorphic robot that is designed to help decrease stress levels and enhance feelings of support and companionship by recreating the sensation of being hugged through haptic interaction” (Levantino, 2018). Sharing heartbeats between people can enhance empathy (Winters et al., 2021). Outside of the scientific literature, there now exist a variety of robotic devices that have been designed to support synchronization between long-distance romantic couples. From a recent review (Lolo Nate, 2022): The Frebble gives the synchronized sensation of holding a partner's hand, the Bond Touch communicates via synchronized tactile feedback, the Lovense supports synchronized sexual stimulation and the Kissenger (kiss messenger) uses actuated silicon lips to replicate the kiss of a distant but synchronized partner.

Embodied robotic movements, like human movements, can communicate emotions. Santos and Egerstedt (2021) found that non-humanoid robot swarms were able to trigger basic emotion perception through simple, basic forms of movement—just modulations of speed and smoothness were able to make robots seem happy, surprised, angry, fearful, disgusted or sad.

Movements create emotive “vitality affects” between infants and parents (Stern et al., 1985). These affects stem from variations in the contours and envelopes of movement intensity and rhythmic patterns. For instance, affective feelings result from motions that are “surging,” “fading away,” “fleeting,” “explosive,” “crescendo,” “decrescendo,” “bursting” “drawn out,” etc. (quoted in Mühlhoff, 2019). Movement-based “body moves” (Gill, 2012) are clearly manifested in robots, such as the non-humanoid toy robots in the popular “furReal” series by Hasbro. For instance, the “Snackin' Sam the Bronto” (Figure 3) toy dinosaur robot moves its neck, mouth and tongue to communicate interest in eating¹. Similarly, “Shybo” (Figure 3), a humanoid machine-like social robot, reacts to loud sounds by turning its hat down and shaking, giving the impression of being scared (Lupetti and Van Mechelen, 2022). A periodic display of facial expressions in a food serving robot (Jiang, 2020) helps contribute to a more friendly vibe—yet, its lack of responsiveness in movement (other than a sudden stop) shows room for future improvement (Figure 3).

Movement-based design improvements have also been applied to non-embodied virtual characters. For instance, Gratch et al. (2021) found that when a digital listener nods and smiles at the right time, people tend to share more information about themselves. The addition of oscillatory motion to the postures of virtual characters tended to increase human empathy in response to virtual expressions of pain (Treal et al., 2021).

This next section progressively builds a theoretical design space (Shaw, 2011; Lomas et al., 2021) to describe the input and output factors of resonance. Table 2 describes how eight different situations emerge from the combination of two factors: the number of participants (i.e., plurality) and their reciprocity. This two factor design space, as an initial gesture, helps reveal different characteristic forms of interactional resonance.

Based on our reviews of resonance in human interactions and in robotics, we then propose additional factors or dimensions to describe the design space of resonant relationships (Box 5). These include the input space, or the independent variables: frequency, amplitude, reciprocity, power balance, plurality, complexity, periodicity, synchrony, predictability, intentionality, fidelity and timescale. The design space also consists of the outcomes, including several objective outcome factors: energy level, frequency, phase, synchronization and stability. Finally, the outcome space also includes subjective outcome:emotional arousal, emotional valence and attentional engagement.

As a research program, we hypothesize that the input factors of resonance can explain subjective and objective outcomes. For instance, the tempo of a robot's interactions (movement and speech) could be systematically varied to determine how this affects the human response. Effects will likely depend on the context (Lim et al., 2021), i.e., they may not always generalize across different robotic platforms, behaviors or cultures.

Researching human resonance may improve human–robot interactions and also help advance human psychology (Sciutti and Sandini, 2017). As people are naturally predisposed to “sync” and “vibe” with each other, this can make the study of their interactions a challenge to scientifically investigate in a controlled manner. Social robots present the possibility of precisely controlling the dynamics of the oscillatory inputs to human social interactions. The paradigm of the Human Dynamic Clamp, for instance, has been specifically proposed to probe the oscillatory relationship between humans and virtual humans (Dumas et al., 2020).

Resonance, as a metaphor, can also help guide scientific research. Bartha (2013) describes how resonance was used as a “programmatic analogy” by nineteenth century physicists investigating spectral lines—the bright lines showing the frequency specific emission of light from molecules. These spectral lines were viewed as “completely analogous to the acoustical situation, with atoms (and/or molecules) serving as oscillators originating or absorbing the vibrations in the manner of resonant tuning forks.” This analogy served as a guiding research program for physicists. Metaphors of resonance might play a similar guiding role in the design of social robots and AI.

This article proposes the possibility of designing autonomous robots that resonate with people at a social level (Henschel et al., 2020). How might roboticists use resonant relationships to improve human–robot interaction quality?

There are many opportunities for robots to support human engagement and collaboration through more oscillatory relationships. Rhythm is recognized as an important non-linguistic cue in Human–Robot Interactions (Mutlu et al., 2016); robotic rhythm may help improve the predictability or legibility of robotic motion (Dragan et al., 2015; Abe et al., 2019). Robots could use their own rhythms to entrain the rhythms of their conversational partner, e.g., by increasing or decreasing the tempo of their conversational interactions. Rhythmic awareness might enable robots to predict when to initiate or cease actions in order to maximize a human response. Social robots could promote interactive social resonance by shaping an appropriate “vibe;” not talking too fast or slow, not talking over other people, not breaking into a conversation at the wrong moment, etc. Alternatively, robots might deliberately interact with existing human oscillations, such as brainwaves, breath, walking, head nodding or heart rate. Robots might gain access to the state of human oscillations through wearable biosensors or they might be able to infer this information from visual or auditory information streams using computer vision or natural language processing. For instance, robots might aim to measure and entrain to the tempo or pace of a person's behavior.

The metaphor of resonance, even apart from physical measurement, may aid designers of social robots and AI systems if the metaphor helps make the complexity of social interactions more manageable. Digital computer interactions involve a great number of metaphors, such as buttons, pipes, folders, files, streams, clouds etc. Metaphors are useful because they provide a conceptual interface between people and a complex system design (Sharp et al., 2019). Resonance may help provide an intuitive model of social interactions that could guide design activities. For instance: although “the vibe” within social groups is far from being understood scientifically, designers of social robots might find the metaphor useful for understanding the reception of social robots.

We suggest that the design vision of “robots that can vibe with people” will lead to distinctly different outcomes than, say, “robots that show empathy.” While the latter might orient toward the mimicry of human facial expressions or the modeling of human emotional states, the former can leverage resonance and vibes as cultural metaphors. This points to subtle visual, auditory and tactile design elements that could be crafted to create emotionally satisfying authentic social interactions.

New opportunities will also arise as we move from resonance as a metaphor to resonance as a mechanism and then to resonance as a measurement. Measures of resonance can play a valuable role in the AI optimization of human experiences; i.e., learning to attune to humans through the maximization of resonance. If resonance can be adequately measured and treated as an metric or objective function, then it might be optimized algorithmically (Lomas et al., 2016). For instance, if interpersonal resonance during a videoconference session could be measured, it could be optimized through the iterative testing of different interventions.

The ability to identify and measure interpersonal synchrony has facilitated a great deal of social research (Condon and Ogston, 1966; Kendon, 1970; Bernieri et al., 1988). Recent efforts have compared different ways of measuring both movement and inter-brain synchrony, using both offline and real-time approaches (Ayrolles et al., 2021; Chen et al., 2021; Dikker et al., 2021; Fujiwara and Yokomitsu, 2021). However, the measurement of resonance presents distinct challenges. Synchrony is simply the statistical correlation of a signal. Measures of resonance may demand more interpretation; taking into account, for instance, the depth of human engagement, its duration, the reverberating echoes of a signal over time, the presence of harmonics, etc. Therefore, it remains a research question: what quantitative metrics might be best matched to the human perception of interpersonal resonance?

To unpack this question, the next sections will consider several of the key factors that are expected to correlate with resonance, including attention, aesthetic pleasure, flow states and wellbeing. During this discussion, hypotheses will be noted with a [H.#] so they can be enumerated in Box 6.

The philosopher Hughes (2012) suggests that robots will need resonance (in the form of a functional equivalent of mirror neurons) in order to demonstrate compassion for people. But, there are negative societal outcomes to consider as well. If robots can resonate with people—that is, build psychological rapport [H.7]—this might significantly enhance their ability to persuade or manipulate people. However, an improved understanding of resonance might also reveal more effective psychological defenses against non-consensual persuasion.

Former US president Donald Trump has been recognized as a political figure with a special ability to resonate with people (Giorgi, 2017). Matheny et al. (2018) provide a close analysis of Trump's acceptance speech at the Republican National Convention. Through an analysis of his body language, they provide “evidence that Trump created an empathetic resonance with the audience that helped generate a sense of political movement and unity.” The authors describe how Trump would characteristically point in the air or put his thumb and finger in a pinch—and then move this gesture in a rhythm synchronized to his own speech rate. The audience cheered 151 times during his speech, 63% of the time during a pinch or pointing gesture. Only 10 of those cheers occurred when Trump was in a bodily neutral position. This analysis shows that affective resonance is a powerful phenomena—but not necessarily a positive phenomena.

Media theorist Gibbs (2019) paints a similarly fraught picture of human resonance at a societal scale:

The systematic application of resonance to political rhetoric at a societal scale may present a deep threat to democracy. How machines, algorithms, or AI might wield resonance as a tool to manipulate humans at scale deserves further study and consideration.

In Box 4, we present an extended hypothesis proposing that animals evolved defense mechanisms to prevent resonance, synchronization and entrainment [H.8]. Similarly, humans seem to have the ability to selectively resonate; one example is that, if we feel that we are being manipulated, we may shut down our openness to resonance [H.9]. Robots that mirror a human user's physical or cultural attributes or express interest in similar ideas or hobbies could potentially enhance the empathic and affiliative response in humans. However, crude “copycat” approaches are likely to easily backfire if people feel manipulated or if they feel the robot interaction is inauthentic (Metzler et al., 2016).

What might happen in a future of resonant robots and virtual agents? There is likely to be a competitive effect where, at first, humans may be compelled but then eventually become more discriminating. People may become used to a higher quality of resonant engagement, which could drive further advances in resonant robots. Eventually, people might become wary of normal levels of interpersonal rhythmic competency making it difficult for normal people to connect. On the other hand, humans may become even more attuned to authenticity in interaction; imperfections and personal character may become more valued.

A theory of resonance offers a bridge or common currency (Northoff et al., 2020) between human conscious experience and the mathematical nature of our physical world. Resonance seems to govern a great deal of neural activity and is also plainly manifest in conscious experience. But, to what extent is the resonance that we feel also the resonance we can observe in the brain? These two domains may only loosely overlap; and yet, our conscious experience must relate, in some way, to the widespread presence of neural oscillations. Hunt and Schooler (2019) have proposed that “the binding problem” of conscious experiences in the mind is achieved through the integration of shared resonances in the brain—in other words, they propose resonance as an answer to the “hard problem” of human consciousness. Subsequent work (Safron, 2020) suggests that Darwinistic competition for resonant amplification underpins consciousness. Valencia and Froese (2020) argue that collective consciousness is shared through inter-brain synchronization.

From an extended mind perspective (Clark and Chalmers, 1998), the rich hierarchy of harmonized oscillations between the brain, the body and external rhythms show how cognitive processing can be distributed into the world (Hutchins, 2001). There may be little difference between an external rhythm and an internal rhythm, other than the degree to which it couples with other neural rhythms. From this perspective, a beating drum is just another rhythm in the brain. Due to the resonant nature of our being, sympathetic resonance with other people seems to allow the direct sharing of collective consciousness across bodies and time. Yet, if resonant coupling is the basis of consciousness, then might we run the risk of creating resonant robots or AI systems that could, in a meaningful sense, actually share our conscious experience?

At present, it is computationally challenging to support real-time resonance with human oscillations (see section Robots That Can Resonate With People or Other Agents). However, there are a variety of new computational architectures that use coupled oscillators to perform information processing (see review by Csaba and Porod, 2020). These naturally resonant computational systems might support a new approach to representing and responding to human activity. Perhaps oscillatory computer systems could become capable of direct resonant coupling with humans in a similar manner to how humans attune to one another [H.10].

This article has presented resonance as a simple physical concept that can explain complex human behavior. While resonance may indeed serve as a powerful program for research, our understanding of it is far from complete. Even the resonance of a stretched string is astonishingly complex (see Bajaj and Johnson, 1992); human resonances, which result from the interrelation of billions of hierarchical oscillators, will no doubt exhibit endless complexities. But, a scientific progression can start with a simple guiding model for resonance that can then lead to a more complex model. As an example, researchers predicted that synchronization might help support romantic courtship behaviors. However, their initially simple notion of synchronization failed to predict behavior. This led the researchers to develop a more complex model of “hierarchically patterned synchronization” that successfully fit the data (Grammer et al., 1998).