---
title: "Ch13 autonomy design content"
description: "Ken Bamforth, a former miner then studying at the Tavistock Institute, goes back down a mine. At the Haighmoor seam he finds something the industry’s own…"
url: "https://wiki.cybernetic-futures.org/punch-and-judy/part-02-outside-the-model/ch13-autonomy-design/ch13-autonomy-design-content"
modified: "2026-11-12T15:26:25.902Z"
language: "en"
site: "Cybernetic Futures"
---

# The Autonomy Design: Trist, Herbst, Cherns, Boeke, Endenburg

Ken Bamforth, a former miner then studying at the Tavistock Institute, goes back down a mine. At the Haighmoor seam he finds something the industry's own engineers had not designed: a set of relatively autonomous groups, interchanging roles and shifts, regulating their affairs with a minimum of supervision.

He brings it to Eric Trist. What they publish in 1951 claims more than that teams are pleasant: for a given technology there is a genuine organizational *choice*, and the design which jointly optimizes the social and the technical beats the one that optimizes the technical alone.

Trist, writing it up thirty years later, is exact about what was seen: "Cooperation between task groups was everywhere in evidence; personal commitment was obvious, absenteeism low, accidents infrequent, productivity high."

That is what an observation looks like when somebody has gone and made it. It is 1951. *Cybernetics and Management* is eight years away and *Brain of the Firm* twenty-one.[^ch13-opener]

***

## The dial that has to be set

The autonomy-cohesion problem↗[^ch13-glossary] is the one Beer addressed most convincingly, and also the one on which the literatures outside cybernetics traveled furthest without him.

The problem is easy to state and hard to escape. Give the parts of a system enough freedom to absorb their own variety↗ and they drift out of alignment with each other and with the whole. Take the freedom back and the center has to absorb variety it cannot possibly hold. Everything Beer built between System One↗ and System Three↗ is an attempt to hold that dial in a defensible position: the principle of autonomy, the anti-oscillation function, the resource bargain, the audit channel that reaches past the manager.

On the requirement, that apparatus is strong. It treats autonomy as a quantity to be allocated, not a virtue to be conferred. The allocation is paid for somewhere, and the bill arrives either as variety the center cannot process or as oscillation between the parts, depending on which way the setting is wrong. Very little management writing gets that far, and it is a large part of why we keep teaching the model.

On the setting itself Beer says more than we teach and less than a designer needs. He claims the setting is determinate: "*Freedom is in principle a computable function of systemic purpose as perceived*," and the metasystem should make "only that degree of intervention that is required to maintain cohesiveness."[^ch13-beer-compute] He assigns the choosing of it to the resource bargain, the "deal" by which "some degree of autonomy is agreed" between senior management and its junior counterparts.[^ch13-beer-bargain] He drafted a questionnaire that asks managers to rate how far the whole and its elements agree on the degree of autonomy, and hoped that "maybe (one day) thresholds will be established."[^ch13-beer-identity] He introduced it with a warning of his own: "it would be bogus to pretend that proper measurements can be made. These are subjective judgments for which we ask."[^ch13-beer-bogus] He did not supply the computation, a validated measure or the thresholds, and nobody in the tradition has supplied them since. Chapter 4, *Channels Without Owners*, made the Part One version of this charge about the channels themselves. The same gap runs through the autonomy question, and restating the principle more emphatically does not close it. We can say that the metasystem, in practice mainly System Three, should spend no more of System One's autonomy than cohesion requires. We cannot say how much that is in a given case, we have no instrument that reports it, and we have not built one.

Three other traditions worked on the same dial. One of them produced a design rule two decades before *Brain of the Firm*. One produced a governance mechanism that a national labor authority will accept as a substitute for a works council. The third assembled an evidence base of a kind cybernetics has never assembled, and then spent its last decade warning people not to use it as a blueprint. That one is Elinor Ostrom's, and Chapter 14 takes it; this chapter takes the first two.

None of these traditions set out to correct us, and two of them appear not to have known we were there. Approached from a coal face, from a design office and from a family electrical engineering firm, the same dial produced components that fit inside our diagram without any help from us, and we did not read them. The loss ran both ways. They worked without the variety vocabulary that would have let them say what they had found; we worked without the field results and the design rules that would have let us specify what we were recommending.

## Trist and Bamforth: the field finding

**What they found ·** A semi-autonomous group inside a jointly designed technical and social whole. **When ·** Haighmoor, 1951. **Who cited whom ·** The lineage predates *Brain of the Firm*. **Evidence base ·** Field observation at one coal seam and a long sociotechnical program. **What it costs us ·** Autonomy becomes a design variable, not merely a permission in a diagram.

Eric Trist and Ken Bamforth published "Some Social and Psychological Consequences of the Longwall Method of Coal-Getting" in *Human Relations* in 1951, twenty-one years before *Brain of the Firm* and eight before *Cybernetics and Management*.[^ch13-longwall] The subtitle does the theoretical work. The paper is about a work group "in Relation to the Social Structure and Technological Content of the Work System", which announces joint determination in the title. What Bamforth had seen at Haighmoor, groups interchanging roles and shifts and regulating their own affairs with almost no supervision, is what Trist summarized thirty years later: "cooperation between task groups was everywhere in evidence; personal commitment was obvious, absenteeism low, accidents infrequent, productivity high."[^ch13-trist81] The Durham comparisons that most people mean when they say "the coal studies" came a decade later, in *Organizational Choice*.[^ch13-orgchoice]

The claim, as the opening of this chapter gave it, is about organizational *choice*. Put in our terms, it says that variety absorption and the machinery cannot be designed separately, and it was made from a coal face rather than from a diagram.

For us the consequence is a change of category rather than a change of belief. Read casually, the finding sounds like something the model already says: of course the operational units need enough freedom to handle their own disturbances. Read as its authors meant it, it says something stronger and less comfortable. The metasystem does not hand the amount of freedom down after the plant has been laid out. Choices made when the technology was specified fix it in large part, and the people who made them were not thinking about autonomy at all. Where a seam is worked by a single interchangeable group, the group can absorb its own variance. Where the same seam is worked by a shift-segmented line, no amount of enlightened supervision restores that capacity. Beer saw part of this. "The loss of autonomy, the erosion of liberty, is already implicit in the total system," he wrote, and it goes unrecognized and gets blamed on autocratic corporate management. But he treated the linkages between operations as given, arising "by happenstance" from the nature of the enterprise.[^ch13-beer-implicit] At a coal face they are a property of the work system, designed in or designed out before anyone draws the diagram.

We can test that claim and have not. It also warns us about our own practice. A viable system diagram drawn over a work system whose technical design has already destroyed the operational units' capacity to absorb variance will be internally consistent and practically empty.

The method can be separated from the finding, and it matters at least as much. Trist and Bamforth went and looked, at a place where the arrangement had emerged without anyone designing it, and then described it precisely enough that other people could argue with the description. The comparison work came later and was slow. Neither step is beyond us; we have simply not taken them often. Our tradition has case reports written by the consultant who did the intervention, a different genre with different failure modes, and very little observation of arrangements that arose without us.

## Herbst and Cherns: the design rule

**What they found ·** Minimum critical specification. **When ·** 1974–1987. **Who cited whom ·** The rule developed inside sociotechnics, independently of a System Three job description. **Evidence base ·** Design principles and case work, without a validated measure of how much specification is critical. **What it costs us ·** The missing instrument leaves over- and under-specification as retrospective judgments.

Philip Herbst wrote "Designing With Minimal Critical Specifications" in 1974, inside the same tradition, and the design rule is his: specify no more than is absolutely essential, and do the work of identifying what is essential.[^ch13-herbst] Two years later Albert Cherns systematized it as the second of nine principles of sociotechnical design. He was scrupulous about the debt, saying the principles "owe more to the writings of others… than to our own originality."[^ch13-mincrit] His formulation is the one to carry: "The negative simply states that no more should be specified than is absolutely essential; the positive requires that we identify what is essential."[^ch13-cherns76]

That is System Three's job description, written three years before *The Heart of Enterprise* and with no knowledge of it. In one respect it also goes further than Beer. The VSM↗ tells System Three what it is *for*: hold cohesion while spending as little of System One's autonomy as possible. Minimum critical specification tells it what to *do*, starting from what is essential.

The two are easy to read as the same thing in different words, but a purpose and a procedure are not interchangeable. Beer did write rules close to Herbst's. The metasystem "*should make minimal use of variety attenuators*" in its dealings downward; you organize a system "only somewhat" and "ride on the dynamics of the system"; the management paperwork is sorted into mandatory red and anti-oscillatory green, so that everyone sees "that a tiny proportion of the total paperwork machinery is red."[^ch13-beer-rules] These rules classify constraints by function. They give no procedure for finding the essential ones. Beer's principle of autonomy is a constraint on outcomes:[^ch13-beer-autonomy] whatever System Three does, the residue of autonomy left to System One should be as large as cohesion allows. Herbst's is a rule of construction. When you write the specification, write the smallest one that closes the essential variances, and treat every additional clause as a cost to be justified. A person drafting a document on a Tuesday can follow the second. The first cannot be followed that way. Practitioners who have tried to apply the principle of autonomy directly will recognize the difficulty; usually the principle is invoked to defend a setting that was arrived at some other way.

Cherns revised the list in 1987, and for a cybernetic reader the revisions are the interesting part. He added *power and authority* and *transitional organization*.[^ch13-cherns87] Those are precisely the two things on which sociotechnical implementations, and VSM implementations after them, most reliably fail. The pattern is familiar enough to be a diagnosis. A structural design is agreed, the units are given nominal discretion, and the authority to commit resources stays where it was. Or a design is agreed that would work once installed, and no one has specified the organization that carries the work from the present arrangement to that one. Cherns did not have our vocabulary for either failure. He had watched both happen often enough to promote them into principles.

There is a further reason to want the rule alongside the principle. The principle of autonomy is stated in a form that makes it very hard to be caught violating. Any restriction on a unit can be defended as required by cohesion, and any grant of discretion as the residue the principle demands, because the principle states the boundary as an equation that nobody has measured. A rule of construction is less accommodating. Of each clause in a specification it asks which essential variance the clause exists to close, and it treats a clause that cannot answer as surplus. A design review can actually put that question, and losing it is part of what the model's silence costs.

The rule has a limit: it remains a rule and not a measurement. Neither Herbst nor Cherns produced a way of telling, in advance, how much specification is critical. Their tradition left that to skilled joint design with the people doing the work, which is a defensible answer and not an instrument. We have no instrument either, and we have less excuse, because the question is stated in exactly the terms our tools were built for.

## What the sociotechnical record will and will not support

The tradition also has evidence, and more of it than we do. In 1982 Pasmore, Francis, Haldeman and Shani reviewed a hundred and thirty-four sociotechnical experiments and reported them successful across productivity, cost, quality and satisfaction. The same review indicts itself: very few of the hundred and thirty-four involved redesigning the *technology*. Most rearranged the social system around fixed plant, which is not what the theory claims to do.[^ch13-pasmore]

That qualification matters more than the headline, and our own literature rarely makes one like it about itself. A hundred and thirty-four cases that mostly tested half the theory do not validate joint optimization. They validate social rearrangement under fixed technical constraint, a narrower and more familiar claim. The authors said so in print, in the same paper that reported the successes.

John Kelly's 1978 reappraisal went further. Autonomy in self-managing groups, he argued, is bounded by economic imperatives rather than genuinely devolved; joint optimization is a myth better described as labor intensification; and pay incentives carry more of the causal weight than the theory admits.[^ch13-kelly] Kelly's argument should be taken as a live hypothesis rather than a historical footnote. It is a rival explanation of the same results, and it applies to us without modification. If devolved discretion in these cases was bounded by economic imperatives set elsewhere, the experiments demonstrated the productivity of a particular reorganization of effort, and not the value of autonomy. Any structural design that claims to devolve discretion while leaving the money and the headcount where they were is exposed to the same reading, and we have not built the study that would distinguish the two.

And then the tradition died without being refuted, which is the part cybernetics should study hardest. The oil shock, the political turn of the eighties, management's preference for control-oriented alternatives and union indifference to shop-floor initiatives all played a part. The decisive cause was an action-research method that could never diffuse beyond its pilot sites. Harry Scarbrough's verdict, quoted approvingly in a recent reassessment of the tradition, is that it became "a kind of intellectual Esperanto: a language championed by a brave and enthusiastic few, but actually spoken by no one else."[^ch13-scarbrough]

That sentence describes a risk we share. The sociotechnical tradition had field results, a design rule, a review of a hundred and thirty-four cases and a serious internal critic, and it still did not diffuse. On each of those counts our position is weaker. Whatever protected the tradition from being wrong did not protect it from being unused, and nothing in our situation suggests we are better placed.

## Endenburg: the mechanism

**What he found ·** Consent, circles and the double link as a structural response to autonomy–cohesion. **When ·** Boeke's consent principle in 1945; Endenburg's later engineering synthesis. **Who cited whom ·** Romme's 1995 scholarly treatment has eight references and no cybernetician. **Evidence base ·** Operating organizations, but no reliable adoption census. **What it costs us ·** The double link answers a mechanism question Beer posed but did not settle.

In the nineteen-seventies, the decade of *Brain of the Firm*, Gerard Endenburg built the sociocratic circle method inside his family's electrical engineering firm. It has four rules. Decisions stand when no member raises an argued objection. Work is done in semi-autonomous circles, each with its own aim, deciding by consent. Adjacent circles are linked by *two* people, a leader appointed downward and a delegate elected upward, both full members of both circles. And people are assigned to functions by consent after open argument.[^ch13-endenburg]

The double link is the sharpest mechanism in this chapter. Beer's recursion↗ requires that command and resource channels between levels have adequate variety and does not say how to build them. Endenburg implements the requirement as a membership rule and makes the upward channel *elected*. That is a variety design and a legitimacy design at once. Beer multiplied the upward channels: six vertical components, an audit that goes directly to the operations, an algedonic bypass. He put the heads of System One, and perhaps "simply workers," into System Five.[^ch13-beer-plural] He never made the upward link an elected membership with a right of objection. The double link adds election and consent to a channel structure Beer already required to be plural.

In our own terms the design repays a closer look. A single link between levels, which is what most organizations have and what our diagrams imply when they are read carelessly, gives the higher circle exactly one account of what is happening below, produced by the person whose performance is being reported. Two links do more than double the bandwidth. The second occupant is selected by the people being described and sits in the higher circle as a full member with the power to withhold consent, so the channel structurally cannot carry only one interested account. Whether that raises the effective variety of the upward channel or mainly its credibility is precisely the question nobody has tested, and the two are different things.

As far as we can establish, the method is also not cybernetic in citation. Every account of its origin traces it to Kees Boeke's consent principle, itself out of the Quaker business method, and to feedback and control as Endenburg had learned them as an engineer.[^ch13-boeke] The earliest scholarly treatment, Georges Romme's in 1995, names no cybernetician at all: not Ashby, not Wiener, not von Foerster, not Beer.[^ch13-romme] This is the same finding as the Galbraith case in Chapter 12, *The Variety Logic*, and the four silences in Chapter 11, *The Nesting: Simon, Koestler, Jaques*, in a fourth place.

Consent is easy to hear as a cultural preference and deserves the same structural reading. The rule that a decision stands unless a member raises an argued objection is a filter on what has to reach the center at all. It admits the objection and discards the preference, and it does so at the level where the objection arises, before aggregation. That makes it a variety-attenuating device in the strict sense, placed where attenuation is cheapest. Whether it attenuates the right things, and what it discards along with the noise, is an empirical matter on which the sociocratic literature is thinner than its confidence would suggest.

Four independent arrivals do not prove that the components are correct. They are evidence that the components are findable, a weaker claim and a more useful one. They also mean that the case for our model cannot rest on having the components, since other people have them. Whatever we are entitled to claim has to rest on the assembly, and Part Four takes up that argument.

## One hard institutional fact

The Dutch Social and Economic Council, the country's tripartite advisory body, will exempt a company from the statutory obligation to establish a works council, and full implementation of the sociocratic circle method is one of the accepted grounds.[^ch13-ser] It is the strongest hard fact in the sociocratic literature, and it is legal rather than empirical. As far as we have been able to find, no model derived from cybernetics has been accepted anywhere as a substitute for statutory co-determination. The same fact also caps the claim: the companies currently holding that exemption on sociocratic grounds number in single figures.[^ch13-serlist]

The exemption matters for a specific reason. It is a judgment, by a body with an interest in employee representation, that a particular structural design delivers what the statute was written to deliver. That is an external assessment against a stated standard, and no VSM implementation has ever had to pass one. It is not evidence that the method performs well on any other dimension, and the small numbers mean it cannot bear weight as evidence of diffusion. What it establishes is that a governance mechanism of this family can be made legible enough, and specified tightly enough, for a regulator to rule on it. Ours has never been specified to that standard.

## The descendants

The descendants are where the evidence gets uncomfortable. Holacracy, formalized by Brian Robertson in 2007, took the circle structure and the name from Koestler's holarchy.[^ch13-holacracy] Zappos adopted it across a workforce of roughly fifteen hundred. Fourteen percent of staff took a 2015 buyout offer, departures reached roughly eighteen percent within a year, and the company moved off the system over the following four.[^ch13-zappos] The reported reason belongs in a cybernetic book. Corporate Rebels reports Tony Hsieh as saying that the circles were still arranged hierarchically, with budget and headcount allocated from the top down, so teams turned inward instead of toward customers.[^ch13-hsieh]

In Beer's exact terms, that is a System Three failure. The governance layer was replaced and the resource channel was not, and the resource channel won. How much this shows needs care. The wording is secondhand, as the note records, and one executive's retrospective account of a reversion is not a diagnosis. What can be said with more confidence is that the observable design put a decentralized decision procedure on top of a centralized allocation of budget and headcount, and that the model predicts a specific consequence for that combination. Units whose discretion does not extend to resources will spend their discretion on the things it does extend to, which are internal. The prediction and the reported outcome agree. Nobody has done the work that would turn an agreement into a finding.

Medium abandoned Holacracy too, and its stated second reason is the sharper one: "the act of codifying responsibilities in explicit detail hindered a proactive attitude."[^ch13-medium] At Medium, Holacracy failed by breaching Cherns's second principle, a rule written more than thirty years before Holacracy existed by the tradition Holacracy's own family tree descends from. Neither community appears to have connected the two.

Taken together, the two cases make a small argument for the design rule rather than against the family of designs. Zappos left the specification too thin in one place, at the resource channel; Medium made it too thick everywhere. Both failures can be described in the vocabulary of minimum critical specification, and both in ours. Neither community had an instrument that would have shown the error while there was still time to correct it. That missing instrument, more than either company's judgment, is the common cause worth pursuing.

## What formal equality does not reach

The best critical study we found looked at a co-housing community through sixteen interviews. It reports real decentralization and reduced coercion alongside four inequalities that formal equality does not touch: founders keep status, members hold divergent expectations of commitment, circles apply the rules inconsistently, and people differ in time, procedural knowledge and emotional skill.[^ch13-king] Beyond that, the peer-reviewed evidence for sociocracy itself is thin, and its own advocates say so.

That is the precise limit. Sociocracy equalizes the decision *procedure* and leaves the distribution of *capacity* exactly where it was. That is a variety problem, stated in a vocabulary that is not ours, and nobody has modeled it. One site and sixteen interviews cannot establish how general the finding is, and a residential community is not a firm. What the study does supply is a well-specified hypothesis: a consent rule redistributes the right to object without redistributing the ability to formulate an objection, and the second distribution determines who is actually heard. Anyone who has watched a well-run participative process converge on the preferences of its most articulate members will recognize the claim. Testing it is within reach of our tools and has not been attempted.

## What the numbers actually say

Earlier drafts of Chapter 10, *Six Components*, claimed that Endenburg's system is installed in more organizations today than the VSM is. No figures support that claim, and the honest position is narrower.

Nobody counts either. Endenburg's own institute claims hundreds of adopters across the nineteen-eighties and nineties and offers no current total.[^ch13-tsg] One sociocratic body reports direct knowledge of about a hundred organizations and assumes there are three times more, an estimate wearing the clothes of a count.[^ch13-sofa] Holacracy publishes an opt-in list and, beside it, a claim of thousands of users with no total.[^ch13-holacracy-directory] Sociocracy 3.0 is openly licensed and by design keeps no registry at all.[^ch13-s3] On our side there is no registry either. The most recent systematic review of the VSM screened seven hundred and sixty-eight documents, found twenty-one studies meeting its inclusion criteria, and concluded that most of what exists is model and simulation rather than implementation, with limited empirical validation and almost no longitudinal work.[^ch13-fathi]

The comparison is between two unaudited numbers, one plausibly in the hundreds and the other, in the documented literature, in the tens. It points one way and it is not a census.

Two things follow. The weaker is that the claim should not be made without a count, and Chapter 10 no longer makes it. The stronger is that the absence is symmetrical and therefore diagnostic. Neither community can say how many organizations are running its design, and neither has treated that as a defect worth fixing. For a tradition whose central claim is that organizations can be designed for viability, not knowing how many designed organizations exist is a peculiar thing to have tolerated for fifty years.

:::: {.as-of}
**As of September 2026**

The numbers in this section move, and one of them moves fast. The Dutch Social and Economic Council's published register of works-council exemptions lists six in total, of which four rest on sociocratic grounds: Talent B.V. and Endenburg Elektrotechniek in 2023, CKO KleurRijk and Schuberg Philis in 2024. The systematic review of the Viable System Model used for the comparison screened seven hundred and sixty-eight Scopus records, eight hundred and thirteen after manual searching and deduplication, and admitted twenty-one studies.

None of these is an adoption count, and neither side of the comparison has one. The register this chapter asks this community to keep does not exist on the other side either. Current values, and the dates they were taken, belong in the companion register rather than in a printed chapter — which is the argument of the chapter that follows this one.
::::

## What this chapter hands forward

**Operationalize minimum critical specification.** Fifty years on there is still no measure of how much specification is critical, and no way to tell over- from under-specification before the fact. That is a variety-engineering question. Cybernetics owns the tools for it and has never built the instrument. Medium over-specified and Zappos left the resource channel where it was; the two failures are different, and anyone who bothers can diagnose both in Beer's vocabulary. The instrument need not be elegant to be useful. A defensible first version would be a procedure for counting the essential variances a specification is supposed to close, and for identifying the clauses that close none of them.

**Test the double link.** Does an elected upward delegate actually raise the metasystem↗'s variety, or mainly confer legitimacy? Either answer would be informative. If the gain is in variety, the double link belongs in our recommended designs as a channel specification. If it is in legitimacy, it belongs there anyway, for a different reason, and our account of what upward channels are for needs revision. The mechanism has been running in real organizations for fifty years without anyone asking.

**Write up Zappos.** It is the cleanest natural experiment this tradition has been handed since Chile: governance decentralized, resource allocation left centralized, and the system reverted for a reason its own chief executive stated in System Three's terms without knowing the name. Nobody has done it. The documentary record is public, the sequence of changes is dated, and the rival explanations can be stated in advance.

## The counter-entry: sociotechnics did not win

Kelly's 1978 criticism that autonomy remained subordinate to economic objectives, and Scarbrough's 1995 obituary, belong in the ledger. They cost the chapter any easy claim that an independently discovered design rule naturally diffuses. A rule can be correct, published, systematized, taught and reviewed across a hundred and thirty-four cases, and still leave almost no trace on how work is organized. If the argument of this part of the book were that good components win, the sociotechnical record would refute it. The argument is narrower: these components were found without us, which tells us something about the components and nothing at all about their prospects.

## The priced receipt

Write the Zappos natural experiment from the existing documentary record: a bounded desk study of the 2017–2020 reversion, governance changes, and the retained resource-allocation channel. Budget ten researcher-days and one legal/source check. Pre-state the rival explanations. If central resource allocation does not explain the reversion better than the alternatives, surrender the claim that the failure was a clean System Three diagnosis. The cheaper study comes first because the documents already exist.

[^ch13-longwall]: E. L. Trist and K. W. Bamforth, "Some Social and Psychological Consequences of the Longwall Method of Coal-Getting: An Examination of the Psychological Situation and Defences of a Work Group in Relation to the Social Structure and Technological Content of the Work System," *Human Relations* 4, no. 1 (1951): 3–38. The dates against which we measure it are Stafford Beer, *Cybernetics and Management* (English Universities Press, 1959), and the first edition of Stafford Beer, *Brain of the Firm* (Allen Lane, 1972); the edition we cite elsewhere in this book is the second, *Brain of the Firm*, 2nd ed. (John Wiley & Sons, 1981).

[^ch13-trist81]: Eric Trist, *The Evolution of Socio-Technical Systems: A Conceptual Framework and an Action Research Program*, Issues in the Quality of Working Life, Occasional Paper No. 2 (Ontario Quality of Working Life Centre, Ontario Ministry of Labour, 1981), 8. Trist's sentence in full reads "Cooperation between task groups was everywhere in evidence; personal commitment was obvious, absenteeism low, accidents infrequent, productivity high." Earlier drafts of this chapter quoted it without the two verbs and with a comma for the semicolon; we have restored Trist's wording. The paragraph immediately before it describes the Haighmoor work organization as "a set of relatively autonomous groups interchanging roles and shifts and regulating their affairs with a minimum of supervision," which is what we paraphrase above.

[^ch13-orgchoice]: E. L. Trist, G. W. Higgin, H. Murray, and A. B. Pollock, *Organizational Choice: Capabilities of Groups at the Coal Face under Changing Technologies; The Loss, Re-Discovery and Transformation of a Work Tradition* (Tavistock Publications, 1963).

[^ch13-herbst]: P. G. Herbst, *Socio-Technical Design: Strategies in Multidisciplinary Research* (Tavistock Publications, 1974). The chapter "Designing with Minimal Critical Specifications" was reprinted in Eric Trist and Hugh Murray, eds., *The Social Engagement of Social Science: A Tavistock Anthology*, vol. 2, *The Socio-Technical Perspective* (University of Pennsylvania Press, 1993), https://doi.org/10.9783/9781512819052-017. Cherns cites Herbst under both forms at once — "Herbst, 1974/Vol. II" — which is one reason the attribution has been easy to lose.

[^ch13-mincrit]: Chapter 10, Six Components, Six Literatures, No Beer, attributes minimum critical specification to the sociotechnical tradition generally, which is right, and this chapter narrows it: the principle is Herbst's, and Cherns named and systematized it while explicitly disclaiming originality. Cherns's sentence, in full, is that the nine principles "represent a distillation of experience and owe more to the writings of others (Emery, Vol. II, 'The Nine-Step Model'; Emery and Trist, 1972; Herbst, 1974/Vol. II, 'Designing With Minimal Critical Specifications') than to our own originality. They have not, however, previously been systematized." The distinction matters only because Cherns is the citation everyone uses.

[^ch13-cherns76]: Albert Cherns, "The Principles of Sociotechnical Design," *Human Relations* 29, no. 8 (1976): 783–92, https://doi.org/10.1177/001872677602900806. Minimal critical specification is Principle 2 of the nine, and the quoted formulation opens Cherns's discussion of it.

[^ch13-cherns87]: Albert Cherns, "Principles of Sociotechnical Design Revisited," *Human Relations* 40, no. 3 (1987): 153–61, https://doi.org/10.1177/001872678704000303. The revised list keeps compatibility, minimal critical specification, information flow, support congruence and incompletion, folds variance control and the multifunctional principle back in, drops *design and human values* as a separate item on the ground that values underwrite all the principles rather than standing beside them, and adds *power and authority* and *transitional organization*.

[^ch13-pasmore]: William Pasmore, Carole Francis, Jeffrey Haldeman, and Abraham Shani, "Sociotechnical Systems: A North American Reflection on Empirical Studies of the Seventies," *Human Relations* 35, no. 12 (1982): 1179–1204, https://doi.org/10.1177/001872678203501207. The self-indictment is the authors' own: sociotechnical experiments have been "extremely successful overall," but "the number of experiments involving technological innovation or change is relatively small," and they concede that "we still have much to learn regarding the design of technical systems for joint optimization."

[^ch13-kelly]: John E. Kelly, "A Reappraisal of Sociotechnical Systems Theory," *Human Relations* 31, no. 12 (1978): 1069–99, https://doi.org/10.1177/001872677803101204. Kelly argues that the autonomy granted to autonomous work groups is "limited, and subordinate to particular economic objectives," that joint optimization "has little connection with the reality of sociotechnical practice" and is better described as labor intensification, and that "the role of pay incentives is also shown to have been seriously underestimated."

[^ch13-scarbrough]: H. Scarbrough, "The Social Engagement of Social Science: A Tavistock Anthology," *Human Relations* 48, no. 1 (1995): 23–33, https://doi.org/10.1177/001872679504800102. We have not been able to confirm the "intellectual Esperanto" sentence against the text of Scarbrough's review-article, which is paywalled, nor to identify with certainty the recent reassessment that quotes it; the sentence circulates in the sociotechnical literature without a page reference. The review-article itself exists and is correctly described. Anyone reprinting the quotation should check it against page proof before doing so.

[^ch13-endenburg]: Gerard Endenburg, *Sociocracy: The Organization of Decision-Making; "No Objection" as the Principle of Sociocracy* (Eburon, 1998). For a compact statement of the four rules see also A. Georges L. Romme and Gerard Endenburg, "Construction Principles and Design Rules in the Case of Circular Design," *Organization Science* 17, no. 2 (2006): 287–97, https://doi.org/10.1287/orsc.1050.0169.

[^ch13-boeke]: Kees Boeke, *Sociocracy: Democracy as It Might Be* (1945), https://www.sociocracy.info/sociocracy-democracy-kees-boeke/.

[^ch13-romme]: A. Georges L. Romme, "The Sociocratic Model of Organizing," *Strategic Change* 4, no. 4 (1995): 209–15, https://doi.org/10.1002/jsc.4240040404. The claim is about citation, not vocabulary: Romme's article does use the word *cybernetics* and does describe circle processes as feedback loops, but it sources that language to Endenburg's own engineering training by way of John Buck and Gerard Endenburg, *The Creative Forces of Self-Organization* (Sociocratisch Centrum, 1984). Its reference list contains eight items and not one cybernetician among them.

[^ch13-ser]: Sociaal-Economische Raad, "Ontheffing van de plicht een or in te stellen," https://www.ser.nl/nl/thema/medezeggenschap/medezeggenschapstructuren/ontheffing. The power is under article 5 of the Wet op de ondernemingsraden. The SER's own wording is that a good ground for exemption may be that the employees are substantially co-owners "of dat de Sociocratische Kringorganisatiemethode volledig is geïmplementeerd" — that the sociocratic circle organization method has been fully implemented. Two qualifications the text does not carry: an exemption is temporary, granted for at most five years and renewable on reapplication, and the SER states that it is sparing with them.

[^ch13-serlist]: SER, page cited in the previous note, section "Overzicht verleende ontheffingen" (consulted 29 September 2026). At the time of writing the SER's published register lists six current exemptions in total, of which four are granted on sociocratic grounds: Talent B.V. (2023), Endenburg Elektrotechniek B.V. (2023), CKO KleurRijk (2024) and Schuberg Philis (2024). The other two — BügelHajema Adviseurs B.V. (2021) and Holland Colours N.V. (2023) — rest on employee self-governance and a works-council-substitute committee rather than on the circle method. "Single figures" is therefore if anything generous.

[^ch13-holacracy]: Brian Robertson, "Organization at the Leading Edge: Introducing Holacracy; Evolving Organization," *Integral Leadership Review*, June 2007, https://integralleadershipreview.com/5328-feature-article-organization-at-the-leading-edge-introducing-holacracy-evolving-organization/; Brian J. Robertson, *Holacracy: The New Management System for a Rapidly Changing World* (Henry Holt, 2015); HolacracyOne, *Holacracy Constitution, Version 5.0* (HolacracyOne, 2019), https://www.holacracy.org/constitution. Robertson credits the term to Arthur Koestler, *The Ghost in the Machine* (Hutchinson, 1967), where the holon is introduced as a whole that is also part of a larger whole.

[^ch13-zappos]: For the headcount and the 2015 offer, Benjamin Snyder, "14% of Zappos' Staff Left after Being Offered Exit Pay," *Fortune*, May 8, 2015, https://fortune.com/2015/05/08/zappos-quit-employees/ — roughly 210 of about 1,500 staff, or 14 per cent, took at least three months' severance by the April 30, 2015 deadline. For the year's total, Jena McGregor, "Zappos Says 18 Percent of the Company Has Left Following Its Radical 'No Bosses' Approach," *Washington Post*, January 14, 2016 — 260 people, 18 per cent, over 2015. See also Bebeto Samantarai and Indu Perepu, *Holacracy at Zappos (B): From Experiment to Exodus* (Case Centre, 2017), https://www.thecasecentre.org/products/view?id=146418. The "following four" years is the interval over which Zappos left the system rather than a single announced exit: the internal marketplace that displaced the circle structure was introduced in March 2017, customer-generated budgeting in September 2018 and the triangle of accountability in early 2019, and the company had quietly stopped describing itself as holacratic by 2020. The three dates are those given in Minnaar, "Zappos's Evolution" (next note). Zappos has never published a statement abandoning Holacracy, and we have found none.

[^ch13-hsieh]: Joost Minnaar, "Zappos's Evolution: From Holacracy to Market-Based Dynamics," Corporate Rebels, October 2, 2021, https://www.corporate-rebels.com/blog/zappos-market-dynamics. The sentence as printed there is "Our circles were still arranged hierarchically, where budgeting or headcount was allocated from the top down." Two cautions. Corporate Rebels gives no interview, transcript or date for the remark, and we could not find it in any Zappos publication, including the Zappos Insights page on Holacracy to which the quotation is elsewhere attributed; Hsieh died in November 2020 and cannot be asked. The gloss that follows — that this turned teams inward rather than toward customers — is Corporate Rebels' summary of the consequence and not Hsieh's own words.

[^ch13-medium]: Andy Doyle, "Management and Organization at Medium," Medium, March 4, 2016, https://medium.com/blog/management-and-organization-at-medium-2228cc9d93e9. Doyle's sentence in full: "we found that the act of codifying responsibilities in explicit detail hindered a proactive attitude and sense of communal ownership." His first reason is coordination cost at scale; the sentence we quote is the second.

[^ch13-king]: Daniel King and Martyn Griffin, "Governing for the Common Good: The Possibilities of Sociocracy in Nonprofit Organizations," *Voluntas: International Journal of Voluntary and Nonprofit Organizations* 35, no. 6 (2024): 1046–57, https://doi.org/10.1007/s11266-023-00627-2. The site is a co-housing community of twenty-five households in the northwestern United States, studied through sixteen interviews and four days of observation including four general circle meetings. The four inequalities are the authors' own categories: status, expectation, application and capacity.

[^ch13-tsg]: The Sociocracy Group, successor to Endenburg's Sociocratisch Centrum, states on its website that over the 1980s and 1990s "hundreds of organizations in the Netherlands and around the world adopted Sociocracy" (The Sociocracy Group, "About Us," https://thesociocracygroup.com/about-us/, checked 29 September 2026). It gives no list, no count and no current total, and neither does Sociocratisch Centrum Nederland. The earliest scholarly description, Romme, "The Sociocratic Model of Organizing," 211, names only a handful of adopters by kind — Endenburg Elektrotechniek, a municipal police department, a consultancy, a chain of hairdressing shops — and adds that the model "has been tested in organizations of up to 1300 people and is being tried out on a limited basis in substantially larger organizations in the Netherlands, Germany, Brazil, Canada, the USA and other countries." That is compatible with hundreds and does not establish it.

[^ch13-holacracy-directory]: HolacracyOne, "Directory," https://www.holacracy.org/directory/, checked 29 September 2026: "This directory lists organizations that choose to be featured," and "Thousands of organizations around the world use Holacracy, but only the ones who've opted in are listed here."

[^ch13-sofa]: Sociocracy For All, "Sociocracy: The Movement," https://www.sociocracyforall.org/sociocracy-the-movement/. The organization states that it "is aware of and connected to about 100 sociocratic organizations" and assumes "there are at least 3 times more organizations that have implemented sociocracy to some significant degree," while conceding that "it is impossible to make a realistic guess of how many organizations are using sociocracy worldwide."

[^ch13-s3]: Bernhard Bockelbrink, James Priest, and Liliana David, *Sociocracy 3.0 — A Practical Guide* (Evolving Collaboration, 2022), https://patterns.sociocracy30.org/. The pattern library is published under a Creative Commons license; there is no adoption registry by design.

[^ch13-fathi]: Mohammad Reza Fathi, Soraya Birami, Alireza Payvar, and SeyedAli Doorafshan, "A Systematic Review of the Viable System Model: Applications, Insights, and Future Directions," *Journal of Systems Thinking in Practice* 4, no. 3 (2025): 109–45, https://doi.org/10.22067/jstinp.2025.91552.1136. The seven hundred and sixty-eight is the Scopus record count; a further forty-five records came from manual searching, for eight hundred and thirteen after deduplication, and twenty-one studies met the inclusion criteria. The authors report that "most studies rely on simulations or conceptual models without real-world testing" and that there is an "absence of longitudinal studies."

[^ch13-opener]: E. L. Trist and K. W. Bamforth, "Some Social and Psychological Consequences of the Longwall Method of Coal-Getting," *Human Relations* 4, no. 1 (1951): 3–38; Eric Trist, *The Evolution of Socio-Technical Systems*, Occasional Paper No. 2 (Ontario Quality of Working Life Centre, 1981), 8, for both quoted passages — the description of the Haighmoor groups and the sentence on what was observed. The dates measured against are Stafford Beer, *Cybernetics and Management* (English Universities Press, 1959), and the first edition of *Brain of the Firm* (Allen Lane, 1972).

[^ch13-glossary]: Terms marked ↗ are defined in the Glossary at the back of the book. The mark is set on the first appearance of each term in each chapter, and not afterwards.

[^ch13-beer-compute]: Stafford Beer, *The Heart of Enterprise* (John Wiley & Sons, 1979), 158; restated at 200.

[^ch13-beer-bargain]: Stafford Beer, *Diagnosing the System for Organizations* (John Wiley & Sons, 1985), 38–39.

[^ch13-beer-identity]: Beer, *Heart of Enterprise*, 417–33 (the "Degree of Autonomy" questions, 424–25), with the sentence quoted at 432.

[^ch13-beer-bogus]: Beer, *Heart of Enterprise*, 419.

[^ch13-beer-autonomy]: Beer, *Heart of Enterprise*, 158, for what Beer calls "the cybernetic argument for autonomy (or freedom)", which is what this book means by the principle of autonomy: the metasystem "should make only that degree of intervention that is required to maintain cohesiveness"; 160, where minimal intervention is "essential to viability, save only that this minimum is compatible with systemic cohesion"; 173, where autonomy is the "institutional synonym" of freedom. Beer does not count it among his Principles of Organization; Chapter 16 uses the same definition.

[^ch13-beer-implicit]: Beer, *Heart of Enterprise*, 152 and 142; Beer, *Diagnosing the System*, 84, where the operational and environmental linkages "absorb whatever variety the circumstances of the enterprise dictate."

[^ch13-beer-rules]: Beer, *Heart of Enterprise*, 157–58 and 180–81; Stafford Beer, *Brain of the Firm*, 2nd ed. (John Wiley & Sons, 1981), 53: "Instead of trying to organize it in full detail, you organize it only somewhat; you then ride on the dynamics of the system in the direction you want to go." See also *Heart of Enterprise*, 340 ("no uniform planning rubric can be imposed") and 344.

[^ch13-beer-plural]: Beer, *Heart of Enterprise*, 143, 263–64 and 407; Beer, *Diagnosing the System*, 82; Stafford Beer, *Beyond Dispute: The Invention of Team Syntegrity* (John Wiley & Sons, 1994), 149.