NOTE-826: Turvey — Preliminaries to a theory of action

Contribution

The essay sets out, in Bernstein’s line and largely after Greene (1971), a set of principles for how coordinated action is produced, and an outline of how vision enters it. It names the units of action “coordinative structures”, and makes tuning, the setting of parameters of autonomous lower units, the means by which higher levels control them. It is explicitly “preliminary and speculative” (p. 2). The empirical content is a review of Soviet and Western reflex-probe studies showing that the spinal segmental apparatus is reorganized before a voluntary movement.

Key insight

The executive cannot know or command the details. There are too many degrees of freedom, and the same impulse to a muscle gives different movements in different conditions. So an act is assembled from units that regulate themselves, and the higher level controls them the way one controls a spring: not by moving it, but by setting its stiffness and its rest length, after which the spring finds its own way to equilibrium whatever the starting point.

Assumptions

• Bernstein’s problem. The relation of innervation to movement is equivocal, and the controlling system has far fewer degrees of freedom than the controlled one (pp. 5–6).
• A state-like executive. Coordinated activity “is under the management of an ‘executive system’” (p. 5), understood as “a set of procedures engaging a number of neuroanatomical structures” (p. 14), not one site.
• Reflexes as a basis. The set of reflexes is assumed to span all acts, as a basis spans a vector space (p. 8, after Easton 1972a).
• Small systems model large ones. “We shall proceed on the assumption that the regulatory principles for large systems follow very much the pattern of small systems” (p. 30).
• Movement-related evidence only. “There are no experiments on segmental pretuning for acts that depend on the detection of environmental properties” (p. 22). The environmental account is extrapolated.

Key results

• Constancy in action (pp. 2–4). The letter A can be written with the fingers, the whole arm, the teeth or the toes. “A required result can be attained by an indefinitely large class of movement patterns”, which cannot be stored, so A-writing is “creative in the sense that language is creative” and rests on abstract, generative procedures.
• Against the keyboard executive (pp. 5–6). Three counts.
  • Reciprocal innervation makes instructing muscles singly “arduous and wasteful”.
  • For one limb segment under one muscle, I·d²α/dt² = F(E, α, dα/dt) + G(α), with E the innervation. If E = E(t, α, dα/dt), “the same impulses may produce different movements and different impulses the same movement”. With a hammer, a baton or a can of beer the same rhythmic movement needs different impulse sequences, which “cannot maintain even a remote correspondence” to the movement (Bernstein).
  • “The number of degrees of freedom of the system controlling action is much less than the number of mechanical degrees of freedom of the controlled system” (Kots et al. 1971). Gunkel: two hands moving at different rhythms modulate each other’s amplitude.
• The coordinative structure (pp. 7–9). Spinal pattern generators fix the form of stepping, and afferent information fits it to the external forces, so a “relatively simple executive instruction” sets the whole segmental apparatus going. “We shall refer to reflexes and functional combinations of reflexes as ‘coordinative structures’ [a term borrowed from Easton (1972a) but used here with greater latitude].” “A closely knit functional combination of reflexes performs as a relatively autonomous unit; by this assumption, relative autonomy is a fundamental property of coordinative structures, whether large or small.” Skill acquisition is “a problem of reducing the degrees of freedom” (fn. 1): in Kots and Syrovegin’s wrist–elbow task the joints moved in one of “three to seven constant ratios”, set by functional, not mechanical, links.
• Action concepts as operators (p. 9). The elements of an action concept are “the operations defined over the set of coordinative structures”; an action concept is “analogous to a mathematical operator”.
• Heterarchy (pp. 10–13). Whether one structure is executive to another is not fixed; “either may treat the other as a relatum, or subprocedure, depending on the problem”. Internal feedback, from spinal centres to cerebellum, checks “predictions of how functional groupings of muscles (coordinative structures) will behave”. An act “evolves as the mapping by a heterarchically organized system of an intention onto successively larger collections of increasingly smaller and less complex coordinative structures”.
• Executive ignorance (pp. 14–15). After Greene, the executive does not know which state a lower level will reach. “The units of information that behave deterministically are not functions but equivalence classes of functions”: “ballparks into which it is sufficient for the executive to transfer the state”.
• The plan as a projection of the environment (pp. 16–17). The action concept of A and the perception concept of A both depend on topological, not metric, properties. Evarts’s motor-cortex units code force, which Turvey reads as a late, “surface” stage. “The action plan unfolds as a series of progressively less abstract projections of the environment.”
• Tuning (pp. 18–22). The γ system sets the equilibrium (static) and damping (dynamic) of the spindle servomechanism, so “the servomechanism is not only informed of what it has done, but … can be informed of what it must do”. Higher-level signals to the γ system and to interneurons act “less as instigators of movement than as modulators of the gain of the feedback loops”. “The control of movement is in many respects the reorganization or tuning of the system of segmental interactions”, and “this attunement precedes the transmission of activating instructions”.
• Evidence of pretuning (pp. 22–26).
  • Gurfinkel et al.: knee extension follows the command at 160–180 msec; a patellar reflex evoked after about the first 100 msec is enhanced, more so the nearer to movement.
  • Gottlieb et al.: the H-reflex is augmented in the 60 msec before plantar flexion and generally inhibited before dorsiflexion.
  • Hip flexion enhances the reflex only of the quadriceps head involved; knee extension enhances both heads; elbow flexion also enhances the patellar reflex, less. “Both specific and nonspecific changes … precede voluntary motor unit activation.”
  • Kots and Zhukov: in the final 30 msec before dorsiflexion reciprocal inhibition of the antagonist is enhanced without any change in its motor neurons’ excitability.
  • The reorganization goes from “initially diffuse” to “more localized” closer to the movement (Gelfand et al.).
• Tuning as parameter specification (pp. 26–30). A spring, F = −S₀(l − λ₀), is controlled by setting S₀ and λ₀. Asatryan and Fel’dman’s unloading experiments show that a joint has an “invariant characteristic” M(φ) fixed by its initial setting, independent of the external load. In Fel’dman’s experiments the arm went to the angle that characteristic defined even when spring and weight were removed unpredictably before movement: “the equilibrium position is independent of the initial conditions”, and correction comes only after the new steady state. Turvey generalizes: coordinative structures are “biasable, self-regulating vibratory systems”, illustrated (not modelled) by m·x″ + k·x′ + s·x = 0, tuned by setting s (stiffness, so equilibrium) and k (damping). In Gottlieb et al.’s data the H-reflex changes do not track the EMG, so “the mechanisms of tuning and activating coordinative structures are largely separate” (p. 30).
• Vision and tuning (pp. 32–38).
  • After Gibson (1958), properties of the optic flow that change control locomotion (outflow to go forward, symmetric expansion for collision, “maximize its optical size” for pursuit); properties that do not change specify what surfaces and objects afford.
  • Tuning reflexes (tonic neck, labyrinthine, eye-muscle stretch) bias the segmental apparatus so that a crude “jump” lands in the direction of gaze. But neck angle is “ambiguous with respect to distance”, so the tuning for a jump’s trajectory must come from optical properties that specify distance and height, “realizable as segmental tunings without the intervention of executive procedures”.
  • “In control reactions the relation is between changing properties of stimulation and patterns of tuning, and in identifying reactions it is between nonchanging properties of stimulation and action plans.” “Different properties of stimulation enter into the unfolding act at different levels.”
• Duality (pp. 40–43). Perceiving and producing may be “dual representations of common neural events”: “the structure that affords perception of a speech sound also affords its production”. The only evidence offered is hybrid crickets, whose females prefer the hybrid males’ song (Hoy and Paul 1973), and Turvey grants the principle may hold “only in the communication mode”.

Claims

id	claim	strength	support
C1	An executive cannot control muscles individually: innervation and movement are equivocally related and the degrees of freedom too many	moderate	Bernstein’s argument, the limb equation and Gunkel’s interference, pp. 5–6
C2	Acts are assembled from relatively autonomous coordinative structures, reflexes and their functional combinations	weak	argued, with the stepping and constant-ratio studies, pp. 7–9
C3	The spinal segmental apparatus is reorganized, specifically and nonspecifically, in the tens of milliseconds before voluntary movement	strong	the reviewed reflex-probe experiments, pp. 22–26
C4	A joint behaves like a spring whose equilibrium is set by parameters fixed in advance and is independent of initial conditions	moderate	Asatryan and Fel’dman’s experiments, pp. 26–29
C5	Tuning and activation of coordinative structures are separately controlled	weak	one dissociation in Gottlieb et al., p. 30
C6	All coordinative structures are tuned as vibratory systems, at every level	weak	extrapolated from the servomechanism, pp. 29–30
C7	Optical properties that change are realized as tunings without executive involvement, those that do not change select action plans	weak	the cat, the mantis and the two-visual-systems evidence, pp. 34–40
C8	Perception and action are dual descriptions of common events	weak	speech-perception argument and one cricket study, pp. 41–43

Concepts

• Coordinative structure. A reflex or a functional combination of reflexes that acts as a relatively autonomous unit; borrowed from Easton (1972a), “used here with greater latitude”. Turvey’s examples are spinal and within one body: stepping, the α–γ servomechanism (“the fundamental coordinative structure”), the locomotor synergism.
• Executive. Not a structure but whichever procedure is, “at the moment, in respect to the task”, acting as executive to another (Greene).
• Heterarchy, coalition. Organization in which the roles of higher and lower are not fixed and domains negotiate.
• Equivalence class (ballpark). The set of states or functions a lower system can interconvert, which the executive specifies instead of a single state.
• Tuning. Biasing the parameters of an autonomous system (gain, equilibrium, damping) without disrupting its self-regulation; movement-related or environment-related.
• Tuning reflex. A reflex whose effect is to bias the segmental apparatus (tonic neck, labyrinthine) rather than to move a limb.

Connections

The essay works from Bernstein’s Co-ordination and Regulation of Movements (1967, LIT-907, held unread), from Greene (1971), from the Gelfand–Tsetlin group’s papers, and from Gibson (1958, 1966). It takes Koestler’s holon (LIT-908, unread) for the whole-part character of substructures and then questions its fixed direction. Its later lineage in the record: Kugler, Kelso and Turvey (1980, LIT-913, unread) recast coordinative structures as dissipative structures; Kelso et al. (1984, LIT-902, unread) supply the jaw-perturbation evidence; and the Haken–Kelso–Bunz line is THEORY-069. Its mass-spring tuning is Fel’dman’s equilibrium-point idea, before that name.

Bearing on the record

• LIT-904’s summary is half right. It says the chapter is “usually credited with introducing the coordinative structure”. Turvey says the term is “borrowed from Easton (1972a) but used here with greater latitude” (p. 8). The chapter is where the term was widened from reflexes to functional combinations of them and made the unit of action control, not where it was coined. The 2026-10-11 curation entry’s “the usual source for the term” should say so.
• The coined sense, for CLAIM-175 and CASE-046. CLAIM-175 records “coordinative structure” among the names discussed for the dancing couple. In Turvey’s sense a coordinative structure is built from reflexes and lies within one body; nothing in the essay extends it across two. What does carry over is relative autonomy, control by tuning, and the executive’s ignorance of the details. CASE-046 calls compensation that preserves the dance “the signature of a coordinative structure in motor control”. That signature, reciprocal compensation, is Kelso et al. 1984 (LIT-902), not this essay. Turvey’s nearest evidence is Kots and Syrovegin’s fixed joint ratios and Fel’dman’s equilibrium that does not depend on initial conditions, which is equifinality, not compensation. An extended use would be the record’s widening, as Turvey’s was a widening of Easton’s.
• THEORY-069. The HKB account models coordination as relative-phase dynamics. This essay is its predecessor in the same programme but models units as tuned springs and says nothing of phase transitions.
• Nucleating-argument line. The executive that specifies only a “ballpark” and leaves its realization to self-regulating parts is the structure of an argument whose owner states a direction and leaves the filing to an agent. This is a reading-across, and the record has no claim it would support or defeat.
• Filed from this reading: THEORY-302, control by tuning of autonomous units, at the strength the reviewed experiments give it for the spinal level and no more.
• No instruction for machine-learning practice; nothing for the anthology.

Limitations

• Preliminary and speculative by the author’s statement (p. 2), with no experiments of its own.
• All pretuning evidence is for simple, movement-related acts; the environment-related account is extrapolation (p. 22).
• The vibratory-system equation is “given only to illustrate a principle” (fn. 7).
• The reading is of the 1975 report; the 1977 chapter may differ.

Open questions

• Whether environment-related pretuning occurs, that is, whether optical information sets segmental parameters before an act without executive involvement; experiments of the reflex-probe kind during visually guided jumping or catching would answer it.
• Whether tuning and activation are separate mechanisms beyond the one dissociation reported.
• Whether the 1977 chapter keeps the attribution to Easton and the examples unchanged.