Contribution
A factorial experiment that separates the two standard explanations of collective synchrony, response to a shared external cue and adaptation between individuals, inside one naturalistic human behaviour. Shared music was switched on or off (same song, or songs 8.5% apart in tempo) and, independently, visual contact between the two dancers (a curtain). Both factors raise synchrony, but in different movement components, and their effects do not interact. The authors also show that the timing of the synchrony does not reveal which factor produced it; its spatial direction does.
Key insight
Synchrony between two bodies is not one quantity. Decompose the movement and different components are synchronized by different causes at the same time: “Within the same dyad, multiple synchronization processes can occur simultaneously, exploiting either the presence of a shared sensory input, or interpersonal adaptation” (Discussion). A measure that sees only timing treats them as one: “Yet examining (movement) temporal features alone, social synchrony might appear as a unitary, indivisible phenomenon, even when it is not.”
Assumptions
- Linear decomposition of posture. The 66-dimensional posture vectors (22 markers × 3 axes), standardised per trial and pooled over all participants, are decomposed by PCA into principal movements (PMs). The first 15 explain over 95% of the variance; an autoencoder did little better (95.6% against 96.5%). Synchrony is then measured per PM, so the result is relative to this basis.
- Synchrony as period locking. Synchrony is cross-wavelet power between the two partners’ PM velocity series, averaged over periods from 0.5 to 16 beats. It is symmetric between the two partners: it says that they are locked, not who follows whom.
- The manipulations isolate the channels. “Music-driven” is the main effect of hearing the same song; “partner-driven” is the main effect of removing the curtain. Partners heard their music through earphones. The paper does not say whether other cues between partners (sound of movement, floor vibration) were controlled; any such channel would be present in both vision conditions and would not affect the contrast.
- Untrained, familiar pairs. 40 dyads were recruited and 5 lost to recording failure, leaving 70 participants; all were laypersons “without dance training”, and “All pairs of participants were familiar with each other”. They stood 2.5 m apart, each in a 0.5 × 0.7 m space, facing each other, without touching.
- Short, structured stimuli. Eight rearranged songs of 20 bars (about 40 s), with instruments entering every four bars; synchrony and the differences between conditions were greatest in the full refrain.
Key results
- Main effects, no interaction. By time-resolved 2 × 2 ANOVAs per PM, Bonferroni-corrected across PMs and cluster-corrected across time, ten of 15 PMs showed a main effect. Partner-driven (visual contact): PM1 side sway, PM3 twist with hands, PM5 side hands, PM11 side shuffle. Music-driven (same music): PM2 vertical hands, PM4 front-back sway, PM8 upper-body rock, PM9 twist, PM14 head bob. Both: PM10 bounce (Figure 2). “we found no evidence of an interaction between musical input and visual contact, and this was so for all PMs.”
- The authors’ reading of independence. “even when participants were presented with non-synchronous music, some PMs would still synchronize with the movements of the dancing partner, just as when participants were presented with synchronous music. Conversely, another group of PMs would keep synchrony with music, even if the partner was visible and moving at a different pace.”
- Geometry. The main axis of the partner-driven PMs was lateral, of the music-driven anteroposterior, of the hybrid vertical (Figure 3A). A second-layer PCA over the ten PMs’ spatial maps, blind to category, gave six eigendirections; four of them (1, 2, 5, 6) predicted the synchrony category “with 100% accuracy” under leave-one-participant-out cross-validation (Figure 4B).
- Timing fails. The same analysis on the periodicity of synchrony gave five eigenperiods that “failed to capture any temporal pattern characteristic of music- or partner-driven synchronies”; the model on the two informative eigenperiods “confused 48% of the data associated with music-driven synchrony with partner-driven synchrony” (Figure 4D). “Hence, the independence between music- and partner-driven synchrony is achieved solely through a spatial, not temporal, organization.”
- Bounce. Bounce alone synchronized through both factors, alone synchronized at the beat (eigenperiod 4), and alone grew in quantity of movement when the partners could see each other (Figure S1).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Shared music and visual contact each raise interpersonal movement synchrony in dyadic dance | strong | 2 × 2 within-dyad design, 35 dyads, cluster-corrected ANOVAs (Figure 2, S2) |
| C2 | They act on different movement components, with one (bounce) affected by both | strong for this sample and decomposition | Figure 2A; depends on the PCA basis |
| C3 | The two effects are independent: no movement shows an interaction | moderate: an absence of evidence, not an equivalence test | Figure S2C |
| C4 | The spatial direction of a movement predicts which factor drives its synchrony | strong within the sample | second-layer PCA and classifier, 100% cross-validated (Figure 4B) |
| C5 | The periodicity of synchrony does not distinguish music-driven from partner-driven synchrony | moderate: a negative result for one temporal representation | eigenperiod PCA and classifier (Figure 4D) |
| C6 | The geometry arises from contextual constraints (lateral moves are visible to a facing partner) or from fixed functional roles of body parts | weak: speculation, stated as such | Discussion |
| C7 | Bounce is a “supramodal pacesetter” of interpersonal synchrony | weak to moderate: three converging observations, mechanism proposed | Figures 2D, 4E, S1; Discussion |
Concepts
- Principal movement (PM): a one-dimensional component of whole-body posture from PCA, “reminiscent of common dance moves”.
- Music-driven, partner-driven, hybrid synchrony: a PM whose interpersonal synchrony shows a main effect of same music, of visual contact, or of both.
- Eigendirection / eigenperiod: a component of a second-layer PCA over the PMs’ spatial maps (body part × axis) or over their synchrony periodicity profiles.
- Geometry of synchrony: the allocation of distinct synchronies to distinct spatial axes and body parts.
Connections
It cites ethology, psychology, musicology and neuroscience for the theoretical distinction between shared input and inter-individual adaptation, and contrasts the two in one behaviour. It builds on the authors’ earlier use of PCA to extract movement primitives from dance and point-light displays (Bigand’s PLmocap). Data are public (IIT Dataverse, doi:10.48557/UR2GBG) and the code is on GitHub.
Bearing on the record
- QUESTION-063 asks what measurement tells a shared drive from coupling. This study runs, in a human dyad, the design the question names: the shared drive and the coupling channel are each removed independently, and the components that follow each are identified. It also gives the negative half of an answer: the timing of synchrony does not discriminate, “when (unlike where) participants move does not reflect whether synchrony is established through music or partner vision.” What it does not supply is the other half of the question’s design, perturbing one member and measuring the other, so direction is not estimated. Filed as THEORY-891.
- CASE-045 (the dancing couple) proposes measuring lead→follow and follow→lead coupling with the music’s beat as an exogenous input. Bigand et al. show the music must indeed be treated as a separate driver, and that it drives some components and not others: a VAR on joint kinematics with the beat as input should expect the pair’s block to be carried by particular components, not the whole body. Their setting differs (no touch, no lead and follow, untrained dancers), and they estimate no direction, no perturbation response and no block gap.
- CLAIM-172 says mutual entrainment marks a combined form of peers, and its 2026-10-14 note that “shared timing alone is not the mark”. This reading supports the latter: two kinds of synchrony, one from a common cause and one from coupling, are indistinguishable in their timing. Visual contact is mutual by design here, but the study does not show that the partner-driven synchrony is mutual entrainment rather than one partner following.
- LIT-933 (Nowak et al.) counts synchrony from mutual influence and from a common external signal as one functional unit, the conflation QUESTION-063 names. This study shows the two coexisting in different components of the same pair, which makes the conflation costly.
- THEORY-520 (participatory sense-making): the partner-driven component is a measured, manipulable instance of movement coordinated through the interaction. Nothing here tests the autonomy of the interaction.
Limitations
- Direction is not measured. Cross-wavelet power is symmetric; nothing says whether one partner led.
- Independence rests on null interactions. The paper reports no interaction; it does not bound one.
- The categories depend on the decomposition. A different basis could split or merge components and move the boundaries between categories.
- Narrow sample and setting. Untrained familiar pairs, face to face at 2.5 m, no physical contact, 40-second songs; the authors say whether the geometry “remains adaptable to changing environmental constraints is yet to be determined”, and that side-by-side dancing might reorganize it.
- The geometric explanation is speculative. Visibility or fixed functional roles of body parts are both offered; neither is tested.
- Effect sizes per movement are not reported in the text, only significant clusters and colour-scaled F-values.
Open questions
- Is the partner-driven synchrony mutual or led? A perturbation of one dancer, or a directional measure on the partner-driven components, would say.
- Does the geometry hold in a partner dance with touch and with lead and follow roles, as in CASE-045? There the haptic channel carries the coupling.
- Does the allocation reorganize when the partners’ relative position changes, as the authors suggest for side-by-side dancing?