5D Library › Embodied Cognition, Bernstein's Degrees-of-Freedom Problem and the Control Theory of Motor Synergies · 中文
Preface: One Action, Endlessly Many Ways
Note
Please read this first. This book is not sports-medicine, physiotherapy or rehabilitation advice, nor a prescription for any sport, instrument or singing training. Anyone with an injury, pain or a neurological condition (for example stroke, Parkinson’s disease or dystonia) should consult a doctor, physiotherapist or other qualified professional, and should not adjust training or rehabilitation on the basis of the models or the checklist in this book.
Apart from published studies that are cited, the arms, muscles, learners, participants and data in this book are fictional or simulated by program, and the parameters are illustrative values, not estimates for any real person or population. “Optimal” in this book means only that, under the stated model and cost function, a method of control or estimation minimises some quantity; the book does not say that any method is “guaranteed” to work or “bound” to improve performance, nor that any program “proves” a theory.
Reaching for a glass of water could hardly look simpler. Mechanically, though, there are endlessly many ways to do it: many combinations of shoulder, elbow and wrist angles bring the hand to the same place, each joint is pulled by several muscles, and the same joint torque can be produced by different combinations of muscles. The Soviet physiologist Nikolai Bernstein turned this into a problem: how does the nervous system choose among so many “redundant” degrees of freedom (Bernstein 1967)? In more than half a century since, motor-control research has offered several answers: “freeze” the extra degrees of freedom; organise them into a few “synergies”; or simply leave alone the variations that do not affect the task and correct only those that do.
This book examines these answers with mathematical models and runnable programs. “Embodied cognition” in the title is used here in a narrow sense: the organisation of movement cannot be understood from the “commands” issued by the brain alone, because the mechanics of the body, the delays of the senses and the structure of the task all help determine what a movement looks like (Wilson 2002; Varela et al. 1991). The book does not discuss the broader claims of embodied cognition about concepts, language or consciousness.
The book asks five questions. First, what exactly is Bernstein’s degrees-of-freedom problem, and what is the difference between calling the extra degrees of freedom “redundancy” and “abundance”? Second, how do uncontrolled-manifold analysis and optimal feedback control describe “controlling only what matters”, and where are the evidence and the limits of these descriptions? Third, can Fitts’s law and the speed–accuracy trade-off be derived from noise and optimality, and how do the minimum-jerk and minimum-variance theories differ? Fourth, how does the nervous system use internal models to overcome sensory delay, and why can you not tickle yourself? Fifth, are muscle synergies “modules” of the nervous system or products of biomechanical and task constraints, and what do studies of “freezing and freeing” in learning and of attentional focus show?
What the book contains
Chapter 1 discusses the origin of Bernstein’s problem, the “redundancy” and “abundance” framings and “repetition without repetition”, and uses a planar three-link arm to show that one task corresponds to a whole set of joint solutions. Chapter 2 discusses uncontrolled-manifold (UCM) analysis and the synergy index, explains normalisation per degree of freedom, coordinate dependence and sampling error, and then demonstrates the “minimal intervention principle” with optimal feedback control. Chapter 3 discusses Fitts’s original and Shannon formulations, the optimised-submovement model of Meyer and colleagues, and the minimum-jerk and minimum-variance trajectory theories. Chapter 4 discusses internal forward models, efference copy and Kalman estimation, the actual range of sensory delays, and the attenuation of self-produced touch and the escalation of force. Chapter 5 discusses the non-negative matrix factorisation of muscle synergies, the problems of VAF thresholds and the dispute over whether synergies are of neural origin. Chapter 6 discusses “freezing and freeing degrees of freedom” in skill learning, the evidence and meta-analyses on external attentional focus, and a model in which instructions act as cost functions. Chapter 7 reads these studies beside Cook Ding carving the ox and Wheelwright Bian in the Zhuangzi and Ji Chang learning archery in the Liezi, and states the boundaries between ancient and modern. Chapter 8 gives a four-step embodied-synergy checklist and a pre-registered randomised controlled trial: in a redundant tracking task where two joints drive a one-dimensional cursor, it compares “external task-outcome focus instructions + redundancy-manifold feedback” with “internal fixed-joint-angle instructions” on transfer accuracy.
The book borders on several other books of 5D Library, which are only briefly recalled here and whose programs are not repeated. Book 10, The Biomechanics of Phonation and Attentional Shift, has already discussed Wulf’s external-focus research, the constrained-action hypothesis, the 2021 meta-analysis of Chua and colleagues and the bias-corrected reanalysis (Chapter 6 only recalls these and adds the transfer outcome and the implications for trial design). Book 29, Vocal Biomechanics, Non-Linear Source-Filter Coupling and Auditory-Motor Feedback, has already discussed the delay of auditory feedback, the pitch-shift reflex and how “internal model + Kalman estimation” damps the oscillation of slow feedback (Chapter 4 deals only with state estimation and does not repeat its control program). Book 34, Psychophysics, the Weber-Fechner Law and Optimal Bayesian Multisensory Integration, has already discussed precision-weighted optimal integration (cited in Chapter 4). Book 23 discusses the difficulty of practice and deliberate practice, and Book 33 publication bias and sample size (cited in Chapters 6 and 8).
Claims corrected after checking
Several statements in the author’s original framework needed correction, supplementation or qualification after checking.
First, Bernstein’s 1967 English volume is The Co-ordination and Regulation of Movements (Oxford: Pergamon Press), a translated collection of his Russian papers (Bernstein 1967). “Repetition without repetition” is quoted in secondary sources from p. 134 and concerns practice; this book did not see the page, and calling it “kinematic modelling” is this book’s extension.
Second, the “principle of abundance” was proposed by Gelfand and Latash in 1998 (Gelfand and Latash 1998) and reviewed in Latash’s 2012 mini-review (Experimental Brain Research 217: 1–5) (Latash 2012). It is a reformulation of the problem, not a measurable “degree of abundance”; the framework’s “丰裕度” is renamed the “principle of abundance”.
Third, the original UCM study was a sit-to-stand task in nine people (eight segment angles) (Scholz and Schöner 1999); variances are compared per degree of freedom, and the usual synergy index is ΔV (Latash et al. 2001, 2002). the original study also reported the ratio V_UCM/V_ORT, but its logarithm log(V_UCM/V_ORT) is not a standard index; this book uses it as an auxiliary index and explains its coordinate dependence (Sternad et al. 2010) and sampling error.
Fourth, a program cannot “prove” optimal feedback control; it can only demonstrate its consequences in a stated model (Chapter 2, ln(V_UCM/V_ORT) = +4.68). The “minimal intervention principle” is given in the original words, “deviations from the average trajectory are corrected only when they interfere with task performance” (Todorov and Jordan 2002); nor is V_UCM ≫ V_ORT a signature unique to optimal feedback control.
Fifth, MT = a + b log₂(2D/W) is Fitts’s original form (Fitts 1954); the Shannon form log₂(A/W + 1) was promoted by MacKenzie (MacKenzie 1992), and the two should not be mixed. The “information-theoretic boundary” is an analogy: the motor system is not a channel of fixed capacity, and the logarithmic form is not the only form that fits the data.
Sixth, minimum jerk (Flash and Hogan 1985, a kinematic smoothness cost) (Flash and Hogan 1985) and minimum variance (Harris and Wolpert 1998, derived from signal-dependent noise) (Harris and Wolpert 1998) are two theories; the velocity profile of the latter changes with limb stiffness (peak from 0.67T to 0.39T).
Seventh, Wolpert, Ghahramani and Jordan 1995 concerns an internal model for sensorimotor integration and does not involve the cerebellum (Wolpert et al. 1995); for cerebellar internal models see Miall and colleagues 1993 and Wolpert, Miall and Kawato 1998 (Miall et al. 1993; Wolpert et al. 1998).
Eighth, “50–200 ms” is only a rough range: the monosynaptic stretch reflex takes about 10–40 ms, the cortical component of the long-latency reflex adds 20–30 ms, and involuntary responses to visual stimuli take about 110–150 ms (Franklin and Wolpert 2011).
Ninth, Blakemore and colleagues 1998 is an fMRI study of six people (Blakemore et al. 1998); the delay-and-rotation tickle psychophysics appeared in 1999 (Blakemore et al. 1999); “you cannot tickle yourself” goes back to Weiskrantz and colleagues 1971 (Weiskrantz et al. 1971). “Efference copy” comes from von Holst and Mittelstaedt (Holst and Mittelstaedt 1950) and “corollary discharge” from Sperry (Sperry 1950); the framework conflates the two terms.
Tenth, d’Avella and colleagues 2003 analysed hind-limb EMG of intact kicking frogs with time-varying synergies and a generalisation of NMF (d’Avella et al. 2003); standard NMF comes from Lee and Seung (Lee and Seung 1999, 2001). Kutch and Valero-Cuevas 2012 state explicitly that they do not disprove the neural synergy hypothesis (Kutch and Valero-Cuevas 2012).
Eleventh, “freezing and freeing” (Vereijken et al. 1992) is not a general law: practice can increase or decrease the dimension of movement (Newell and Vaillancourt 2001), and in a cross-sectional comparison of violin learners, more practice went with less shoulder motion (Konczak et al. 2009).
Twelfth, external focus: Wulf’s 2013 review (Wulf 2013); the 2021 meta-analysis of Chua and colleagues (performance g = 0.264, retention 0.583, transfer 0.584) (Chua et al. 2021); 0.01–0.15 after correction for reporting bias by McKay and colleagues, with Bayes factors favouring a null effect throughout (McKay et al. 2024). The latter was published in Psychological Bulletin in 2024; the figures are from version 3 of its preprint.
Thirteenth, the subtitle keeps “skill advancing into the Way” (技进乎道), a concise formula in common later use; the received text of “Nourishing the Lord of Life” in the Zhuangzi reads “what I care about is the Way, which goes beyond skill” (臣之所好者道也,进乎技矣) (郭庆藩 撰,王孝鱼 点校 1961). The other lines quoted from “Nourishing the Lord of Life” and “Way of Heaven” agree with the received text (for 斫 the edition followed writes 斲); in “Nourishing the Lord of Life” the ctext edition writes 譆 for 嘻; Ji Chang in the Liezi has two e-text variants, noted in Chapter 7.
Fourteenth, the framework asks to compare “transfer accuracy and log(V_UCM/V_ORT)”; this book sets transfer accuracy as the single primary outcome and the log ratio as secondary, because the synergy index can change greatly while task error stays the same (Chapter 6).
Each chapter states its corrections, and they are summarised in the accompanying editor’s notes.
How to read the boxes
As in the other books of 5D Library, different kinds of statements are marked separately.
Evidence
Evidence boxes report published research results with their sources.
Calculation
Calculation boxes give formulas and short Python programs. Every program runs unchanged in a standard Python 3 environment, uses only the standard library and fixed random seeds, and is followed by its actual output. The arms, muscles, learners and data in the book are fictional or synthesised by program, and the parameters are illustrative.
Hypothesis · not established
Hypothesis boxes are the author’s ideas, marked “not yet established”, with a statement of what result would show them to be wrong.
Source text
Source boxes quote classical texts with the author’s plain-language explanation.
Note
Note boxes give reminders and explain choices.
About this edition
This is version 1.0.0, the first complete edition: preface and eight chapters, in Chinese and English. The book contains 14 programs and 5 vector figures. The text, boxes and code were drafted with an AI writing tool from the author’s title, framework and notes; the author is responsible for reviewing and revising them and for checking every source. Corrections and missing references are welcome through the 5D Library website.
Disclosure
The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and therefore has a commercial interest in the subject of this book. The author teaches singing at vocal.guru and singingasia.com and develops movement and perception training courses, and has a direct commercial interest in the subject of this book. The book promotes no course, card deck or service and is not sports-medicine, physiotherapy or rehabilitation advice; the Chapter 8 trial is designed to be run and analysed by researchers independent of the author’s business and does not recruit from the author’s community, students or customers; all the code and the study protocol are published openly for anyone to check.
References
Fan, Y. (2026). Embodied Cognition, Bernstein's Degrees-of-Freedom Problem and the Control Theory of Motor Synergies: From the Uncontrolled Manifold Hypothesis and Internal Forward Models to Fitts's Law, Read Beside Classical Chinese Thought on Skill and the Way [具身认知、伯恩斯坦自由度问题与运动协同的控制理论] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828057 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).