Dance training improves the CNS's ability to utilize the redundant degrees of freedom of the whole body

The study examines how the central nervous system coordinates multiple body segments to maintain balance during external perturbations, focusing on differences between professional dancers and non-dancers. Motor synergies, defined as task-specific organizations of redundant degrees of freedom (DoFs), can stabilize performance variables such as the whole-body center of mass (CoM). Using the Uncontrolled Manifold (UCM) approach, variability can be partitioned into task-relevant (affecting performance) and task-irrelevant (not affecting performance) components; a higher ratio of task-irrelevant to task-relevant variability indicates better exploitation of redundancy. Prior evidence shows healthy individuals typically exhibit greater task-irrelevant variability, which aids adaptation to perturbations, fatigue, and dual-tasking, while pathological/older groups often show reduced task-irrelevant variability. Given dancers’ specialized training and superior balance, the authors hypothesized that dancers would show superior coordination—especially under unstable conditions—manifested as smaller task-relevant variability and greater task-irrelevant variability when responding to a mechanical perturbation. They employed the extrapolated center of mass (xCoM) to account for dynamic stability and analyzed two sequential phases: a more stable phase (xCoM inside base of support, BoS) and a less stable phase (xCoM outside BoS).

The paper reviews the UCM framework as a method to assess how redundant DoFs are organized for task stabilization, with evidence across multiple tasks that healthy individuals often show greater task-irrelevant than task-relevant variability, conferring flexibility to cope with perturbations, fatigue, and secondary tasks. In contrast, pathological and older populations exhibit altered synergies (reduced task-irrelevant and increased task-relevant variability). The authors highlight the concept of dynamic postural control via xCoM derived from the linear inverted pendulum model, noting its relevance to dynamic tasks over static CoM measures. Prior studies suggest dancers possess superior balance in challenging conditions (e.g., one-legged stance, balance on a moving platform) and that experts in various skills often exploit redundancy via increased task-irrelevant variability without performance loss. However, how dancers coordinate multi-segment actions to stabilize whole-body CoM in dynamic conditions had remained largely unexplored.

Participants: Twenty healthy females participated: 10 dancers (age 27 ± 1.89 years; height 161.70 ± 2.95 cm; weight 49.85 ± 3.20 kg; professional training 15.8 ± 4.26 years) and 10 non-dancers (age 23.8 ± 3.79 years; height 161.41 ± 3.23 cm; weight 51.11 ± 4.13 kg). Exclusion: no vestibular or lower limb orthopedic injuries in the prior year. Ethics approval by Hanyang University; informed consent obtained.

Experimental setup: Nineteen reflective markers were placed bilaterally on acromion, lateral elbow, lateral wrist, third-finger tip, greater trochanter, lateral femoral epicondyle, lateral malleolus, heel, toe, and head vertex. Participants stood barefoot with hip-width stance, wearing a waist belt attached to a custom pulling apparatus. A motor-induced pulling impulse (anterior direction) was delivered at random time within a 5-s window. Spring stiffness was tuned; 20 N/cm produced one forward step (not two) for all participants. Kinematics were captured with six infrared cameras (Visol Inc.) using Kwon3D XP at 100 Hz. Data were processed in MATLAB.

Variables and computations:

• Extrapolated CoM (xCoM): xCoM = CoM + CoṀ/ω0, with ω0 = sqrt(g/l), l being distance from lateral malleolus to whole-body CoM. Whole-body CoM position and velocity were computed as weighted sums of 14 segment CoMs and velocities (head, trunk, bilateral upper arms, forearms, hands, thighs, shanks, feet) in AP and ML directions using anthropometric segment masses. This yields xCoM[t] = A · ixCoM[t], where ixCoM[t] are segmental extrapolated CoMs relative to perturbation onset (t = 0).
• Elemental variables for UCM: To ensure independence, relative segmental xCoMs (iRxCoM) were defined within five serial chains branching from the trunk (base): head, right/left arms (upper arm–forearm–hand), right/left legs (thigh–shank–foot). The performance variable was whole-body xCoM.

UCM analysis:

• Task equation: xCoM[t] = f(iRxCoM[t]) with a time-invariant Jacobian J. The null space of J defines task-irrelevant space; its orthogonal complement is task-relevant space.
• Projection: iRxCoM deviations (relative to iRxCoM[0]) were projected to task-irrelevant (iRxCOM_TIR[t]) and task-relevant components (iRxCOM_TR[t]). Phase-specific baselines were used: for Phase I baseline at perturbation onset; for Phase II baseline at the moment xCoM(AP) crossed outside the contralateral foot’s forward BoS edge.
• Mean squared deviations (MSD): Computed over N samples within each phase for task-relevant (MSD_TR) and task-irrelevant (MSD_TIR) components as the sum of segment-wise MSDs. The synergy index, SYN = (MSD_TIR / MSD_TR) * ((n - d)/d) * (1/(MSD_TIR + MSD_TR)), with n = 14 elemental variables and d = 1 performance variable, quantified the extent to which segmental variability stabilized xCoM. Segment contributions (iMSD_TR, iMSD_TIR, iSYN) were also computed.

Phases:

• Phase I (stable): from perturbation onset to when xCoM(AP) exited the contralateral BoS.
• Phase II (unstable): from BoS exit to ipsilateral foot-ground contact. BoS forward edge used the actual toe position estimated from heel-to-toe vector scaled by foot size.

Statistical analysis: Two-way repeated-measures ANOVAs with factors Group (dancers vs non-dancers) and Segment (14 segments), performed separately for Phase I and Phase II and for AP and ML directions, with Bonferroni-corrected post hoc tests. Alpha = 0.05. Data reported as mean ± SE.

• Stepping foot distribution: Across participants, 7 used the left foot and 13 the right foot for the forward step. Stepping side defined as ipsilateral (IS), non-stepping as contralateral (CS).
• ML direction: No significant differences between dancers and non-dancers in either Phase I or Phase II for any dependent measures.
• AP direction, Phase I (xCoM inside BoS): No significant group differences in MSD_TR, MSD_TIR, SYN, or segment-wise iMSD_TR, iMSD_TIR, iSYN, indicating similar postural responses early after perturbation.
• AP direction, Phase II (xCoM outside BoS, unstable phase): • MSD_TIR: Significantly greater in dancers than non-dancers (F1,18 = 5.986; p = 0.025), with a significant Group × Segment interaction (F1,18 = 2.399; p = 0.005). Segment-wise differences contributing to higher MSD_TIR in dancers: head (p = 0.013), trunk (p = 0.044), CS thigh (p = 0.021), IS shank (p = 0.011). No significant differences at CS shank (p = 0.083), CS foot (p = 0.961), IS thigh (p = 0.302), IS foot (p = 0.067), CS upper arm (p = 0.076), CS forearm (p = 0.183), CS hand (p = 0.281), IS upper arm (p = 0.117), IS forearm (p = 0.245), IS hand (p = 0.268). • MSD_TR: No significant group difference (F1,18 = 3.700; p = 0.070); no segment-wise differences in iMSD_TR. • Synergy index (SYN): Significantly greater in dancers (F1,18 = 5.152; p = 0.036), with Group × Segment interaction (F1,18 = 3.069; p < 0.001). Segment-wise SYN higher in dancers at head (p = 0.013), CS thigh (p = 0.023), IS shank (p = 0.011); no differences at trunk (p = 0.075), CS shank (p = 0.084), IS thigh (p = 0.597), upper limb segments on either side (all p > 0.08).

The results support the hypothesis that dancers better exploit redundant degrees of freedom to stabilize whole-body dynamics under unstable conditions following a perturbation. In the AP direction during Phase II, dancers displayed greater task-irrelevant variability (MSD_TIR) without increases in task-relevant variability (MSD_TR), yielding a higher synergy index (SYN). This indicates that dancers reorganize segmental configurations in motor-equivalent ways to keep the whole-body xCoM trajectory stable, rather than reducing individual segment variability. Segment-level analyses implicated the head, trunk, contralateral thigh (stance leg/hip strategy), and ipsilateral shank (swing leg/ankle positioning) as key contributors to dancers’ enhanced synergy patterns, aligning with prior evidence that head-trunk coordination stabilizes head motion and that ankle and hip joints are central to corrective actions and foot placement control. The absence of group differences in Phase I and in the ML direction suggests that the superior coordination of dancers is most evident during the more destabilizing, task-critical phase in the sagittal plane. These findings are consistent with expertise literature showing experts in other domains utilize greater task-irrelevant variability while maintaining performance. The study extends understanding of CNS control by demonstrating that long-term dance training is associated with enhanced capacity to harness redundant DoFs across the whole body during dynamic postural control.

This study introduces a within-trial UCM-based approach using relative segmental extrapolated CoMs to quantify multi-segment kinematic synergies during perturbation-evoked stepping. Professional dancers exhibited greater task-irrelevant variability and a higher synergy index in the unstable phase (xCoM outside BoS) without increased task-relevant deviation, indicating superior exploitation of redundant degrees of freedom to stabilize whole-body CoM. These results suggest that prolonged, specialized dance training enhances the CNS’s flexibility in coordinating multi-segment actions for dynamic postural control. Future research should investigate trial-to-trial variability alongside within-trial measures, incorporate segment orientation information, and explore generalization across tasks and populations to further elucidate CNS control mechanisms.

• Elemental variables were segment CoM positions (and their velocity-adjusted forms) without explicit segment orientation information, which may contribute to coordination during movements involving significant rotations (e.g., turning/twisting), though likely less critical in the current sagittal-step task.
• Estimations relied on anthropometric models (segment masses, inertial properties), propagating uncertainty to whole-body CoM/xCoM calculations. While arguably more direct than joint angle-based approaches (which also require uncertain joint center estimates), these assumptions still introduce error.
• The within-trial UCM approach differs from conventional trial-to-trial linearizations using Jacobians around mean trajectories; each captures distinct control features. The current results thus pertain to within-trial variability and may not directly generalize to trial-to-trial analyses.