Punch and Judy — Overview

5 min read

Punch and Judy — a stress test for the Viable System Model

This source set is the wiki’s presentation of Punch and Judy: A Stress Test for the Viable System Model by Krishan Mathis and Margeret Heath (Tautai Press), a v0.5.0 draft of 2026-08-16. It is written for the people who teach the model, apply it in organizations and argue about it at conferences — the authors are two of them, and the book’s first-person plural is meant literally.

Where the Pure VSM manuscript establishes what Beer’s VSM says, this book examines it. Together the two source sets are the wiki’s two imported manuscripts: one states the model faithfully, the other asks what can be held against the claim that it works.

The charge

The book opens on a single narrow, serious charge: the Viable System Model has never been put at risk. Quantitative tests have been published, and the book examines the two most substantial. Its reading is that neither was built so that a negative result could cost the model its central claim. So the model has not been shown false — and not been shown right either, because the community around it has never stated what observation would make it give something up. The prologue’s own summary of the difficulty:

A framework that can absorb any outcome is a language, not a theory.

The prologue also records Beer’s own commitment, from Brain of the Firm (2nd ed., 1981, 377), that his models “are cheerfully submitted to Popperian criteria of disproof” — and notes that he did not say which observation would count.

The book is explicitly not an obituary, not a defence, and not a demolition. It is a test by people inside the tradition, and it is written so that it could come out empty.

What it does, part by part

Part I — Inside: the model we did not read (Chapters 1–9). Nine things the authors inherited and did not verify. Three jobs are distinguished. Chapters 1–3 set the authors’ teaching beside Beer’s text — the neuroanatomical mapping, System Four’s narrowing to environmental scanning, and the law of requisite variety as it is quoted (including a verb Beer changed between 1966 and 1979). Chapters 4–7 take places where Beer stated a principle, requirement or intention that neither he nor the practice made operational — the channels between recursion levels, the criterion for identifying levels, people in the notation, and an anatomy of maintenance with no account of becoming. Chapter 8 reads the founding case record (Project Cybersyn in Chile). Chapter 9 asks why none of the first eight has ever cost the tradition anything, and organises the rest of the book.

Part II — Outside: the people who found it without us (Chapters 10–16). The longest and most technical part, about a third of the book. Its stance is explicitly that of a debtor, not a competitor: organization design, hierarchy theory, ecology, sociotechnical systems design and sociocracy, the study of the commons, and holonic manufacturing each converged on components the model treats as its own — and, in most cases, measured and published them. The part closes with the three published objections the book takes to land.

Part III — False friends and enemies (Chapters 17–22). If the defects never cost anything, who was not charging for them? The first half is about communities that appear to be allies: neighbours whose support is real, but whose epistemic settlement, once adopted, made the model unchargeable. The book’s line is that a criterion adequate to an instrumental claim — does this diagnosis help? — cannot decide a constitutive one: that there are five subsystems, that they are necessary, that they are sufficient. The second half treats the adversaries and finds them fewer and weaker than the reception suggests, with the expensive one not being a person. The part is candid that not everything in these six chapters has been verified, and says so in the open.

Part IV — What we lose (Chapters 23–27). Part Two conceded almost everything, so what remains? The answer is a sentence: that the model is the only available instrument that will take a named system at a named level and say what that system needs in order to remain viable. Five chapters defend it, and the part is governed by one rule stated as a rule — every credit carries its concession in the same section. The part is written so that it could come out empty, and says what should happen if it does.

Part V — What’s at Risk in the VSM (Chapters 28–30). The receipts attached to the charges in Parts One to Three are collected, ordered by what they would tell us per unit of effort, and made into a priced programme: what observation would make the model give something up, the two decisive experiments that could be run now with existing data and access, and what the whole programme costs. The book states that the experiments are designed so that they could be lost, and that a negative outcome is a publishable finding rather than a failure.

Coda — Judy. The book’s last turn: a cartel can be exemplarily viable, and the model cannot tell you what to point it at. Naming the system is the most consequential act in an engagement and the one act the method names and leaves without a procedure. The coda sets out what it calls the standing test and closes with a small ask to practitioners.

Status and how to read it

  • Draft, v0.5.0 (2026-08-16). Not project canon, and not published as a book.
  • The critique is the authors’ claim, not an established result. These pages never present it as a verdict on the model; they report what the manuscript argues.
  • Where the manuscript marks a finding provisional, records a negative search result, or says an item is unverified, the chapter pages mark it too.
  • Locators are Beer’s own books and papers with printed pages, at the locators the manuscript’s own notes give. They are reproduced as given, not re-derived here.