Contribution
A compact, first-person statement, thirty-six years on, of what the HKB model is, how it was extended, and what research it seeded. For this record its contribution is that it puts the model’s equations, regimes and extensions in one readable place, standing in for the paywalled 1985 paper and the 1995 book.
Key insight
A system of many interacting parts can be described, near an instability, by one or a few collective variables whose dynamics is low-dimensional. For two hands that variable is relative phase. Its landscape has two wells at slow speeds and one at high speeds; the loss of a well is a phase transition, announced by growing fluctuations and slowing recovery. The same landscape, tilted by a difference between the parts, has no wells at all but still bends the trajectory: metastability, where parts both coordinate and keep their own tendencies.
Assumptions
- Order-parameter description near instabilities, after Haken’s synergetics: fast components are “enslaved” to slow collective variables (§2).
- Original HKB (Box 1, top row): spatial and temporal symmetry, identical components (Δω = 0).
- Coupled hybrid oscillators (Eqs. 4–6): each limb ẍ + f(x, ẋ) with f = αẋ + βẋ³ + γẋx² + ω²x, coupled through (ẋ₁ − ẋ₂){A + B(x₁ − x₂)²}; this yields Eq. (1) with a = −(α + 2βR²) and b = ½βR².
- Movement frequency acts as the control parameter through b/a.
Key results
- Eq. (1)–(2). dφ/dt = −a sin φ − 2b sin 2φ; V(φ) = −a cos φ − b cos 2φ. Fixed points at φ = 0 and φ = π for a, b > 0. Bistable for b/a > 1/4; for b/a < 1/4 φ = π is unstable and only in-phase remains. Pitchfork bifurcation.
- Eq. (3), stochastic HKB (Schöner et al. 1986): additive Gaussian noise of strength Q. Fitted means, variances, relaxation times and switching times “with only a single free parameter” (§3.1, as reported).
- Eq. (7)–(8), symmetry breaking: dφ/dt = Δω − a sin φ − 2b sin 2φ, Δω ≈ ω₁ − ω₂. Switching becomes a saddle-node bifurcation; beyond it no fixed points remain but the potential keeps curvature: metastability.
- Eq. (9)–(10), generalized HKB: Kuramoto phase oscillators with HKB’s second-order coupling added: dφ_i/dt = ω_i − Σ_j a_ij sin(φ_i − φ_j) − …, with a_ij, b_ij > 0 the two coupling strengths. The text layer truncates Eq. (10) after the first sum; the paper describes it as “a combination of Eqs. 7 and 9”. Without the second-order term, Kuramoto “cannot handle the presence of both in-phase and anti-phase coordination and transitions between them” (Appendix).
- Kuramoto versus HKB (§3.3). Kuramoto’s transition is disorder to order; HKB’s is order to order, or order to metastable partial order.
- Brain (§4). Reported: phase transitions in MEG and EEG; activation of premotor, insula, pre-SMA and cerebellum scaling with the instability of anti-phase but not changing for in-phase at the same rate (Jantzen et al. 2009); TMS switching anti-phase to in-phase only (Meyer-Lindenberg et al. 2002).
- Generality (§3.2). The same multistability, transitions and metastability are reported within a person, between person and stimulus, between people, across species, and with virtual partners.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Bimanual anti-phase coordination loses stability at a critical frequency and switches to in-phase, as Eq. (1) predicts | strong as reported | the 1985 model plus Kelso’s experiments and Schöner et al. 1986, cited, not read here |
| C2 | The switch is a nonequilibrium phase transition (critical fluctuations, critical slowing) | moderate here | stochastic theory and data cited; not shown in this paper |
| C3 | Symmetry-broken HKB’s metastable regime is the theory of relative coordination | moderate | Eq. (7) analysis; conceptual identification with von Holst |
| C4 | One coordination dynamics governs within-person, between-person and human–machine coordination, so coupling is informational | weak to moderate | breadth of cited findings; the inference to “informational” is the author’s |
| C5 | Brain networks are recruited according to the stability of coordination patterns | moderate as reported | Jantzen et al. 2009, Meyer-Lindenberg et al. 2002, cited |
| C6 | Deriving HKB from neural populations settles whether it is mechanistic | weak | assertion, §6 |
Concepts
- Order parameter / collective variable. A low-dimensional relational quantity (here φ) carrying the information needed to describe the system’s coordination.
- Control parameter. A parameter that, varied, moves the system through its patterns without specifying them (movement frequency); not an independent variable in the usual sense.
- Intrinsic dynamics. The coordination tendencies a system has before learning or instruction.
- Functional information. Meaningful information (intention, memory, perception) acting as an attractor in the same space as the intrinsic dynamics.
- Metastability. Coordination with no stable states, only “trapping” and “wrapping” tendencies.
Connections
Its physics is Haken’s synergetics; its biological problem is Bernstein’s degrees-of-freedom problem and Pattee’s origin of coordination. Kuramoto’s model shares the synergetics origin (§3.3). Its descendants named in §5 include task dynamics, articulatory phonology, ecological dynamics and the excitator model. De Jaegher & Di Paolo (LIT-571) build on the same absolute/relative coordination from Kelso 1995. Raja et al. (LIT-598) use the HKB explanation of coupled oscillators as their example of a dynamics already explained without a Markov blanket.
Bearing on the record
- Behavioral integration. A quantitative case of many effectors acting
as one through a collective variable rather than a central controller.
The record’s other
behavioral-integrationholdings are about control by competition, hierarchy or evidence accumulation; this is the dynamical alternative. - THEORY filed as THEORY-069 (Proposed until the 1985 paper and Schöner et al. 1986 are read): in human rhythmic bimanual coordination, anti-phase loses stability at a critical movement frequency and switches to in-phase through a nonequilibrium phase transition, captured by dφ/dt = −a sin φ − 2b sin 2φ, with critical fluctuations and slowing as early signs.
- Markov blankets (LIT-598, LIT-526). HKB is the record’s worked example of explanation by a collective variable, the alternative Raja et al. set against the blanket partition.
- No instruction for machine-learning practice.
Limitations
- Retrospective and selective by its own account (“not a review”); the evidence for every empirical claim is in cited papers not read here.
- The claim that deriving HKB from lower levels settles the mechanism debate is asserted, not argued.
- The generalization to “mind” (§7) is suggestive language, e.g. Sheets- Johnstone’s “ideational phase transitions”.
Open questions
- Read Schöner, Haken & Kelso (1986) and the 1985 paper to promote C1–C2 from “as reported” to read.
- Is the “informational” coupling claim (C4) distinguishable from mechanical or perceptual coupling in any experiment the paper cites?