---
title: "content"
description: "Source: Stafford Beer, Management Science: The Business Use of Operations Research (MS 1967). Quotations from MS were checked against the page images of the…"
url: "https://wiki.cybernetic-futures.org/pure-vsm/part-05-method-and-epistemology/ch27-management-science/ch27-management-science-content"
modified: "2026-11-12T15:26:26.143Z"
language: "en"
site: "Cybernetic Futures"
---

# content

*Source: Stafford Beer, Management Science: The Business Use of Operations Research (MS 1967). Quotations from MS were checked against the page images of the Internet Archive lending copy and are taken from the reading notes in **`_work/ms-1967-reading-notes.md`**; no searchable full text of MS is available. Quotations from other books by Beer were checked against the full texts.*
In 1967 Beer gave the sixth chapter of a short illustrated book the title "The Viable Governor". The governor in that chapter is first the Watt governor of the steam engine, which brings its variables back into control in the act of going out of control, and then a population of interacting animals, which governs itself without any governor at all. *Management Science* is the shortest and most openly popular of Beer's books from the 1960s. It appeared in a series for a general readership one year after *Decision and Control* (DC 1966) and covers much of the same ground without formulae: operational research (OR), models, probability and the cybernetics of control. Its seven chapters run from the manager's tasks, through chance and variety, models and simulation, and two worked cases, to the chapter on the governor and a closing chapter on automation.
The book shows in plain language how Beer joined OR and cybernetics before *Brain of the Firm* (1972). It also contains, scattered across its chapters, several ideas that the later books assemble into the Viable System Model: requisite variety, intrinsic control, homeostasis and ultrastability, coupled homeostats with preferred states, a hierarchy of models, identity, and the word *viable* itself. Whether and how MS relates to the VSM is a question for the last section, and there only as far as the texts allow.
## 27.1 Management, science and the underlying system
MS opens with the manager's tasks: "Policy-making, decision-taking, and control: These are the three functions of management that have intellectual content" (MS 1967: 10). The book is organised around these three functions. The last, control, becomes the subject of its chapter 6, where the steam governor serves as the model (27.6–27.11 below).
Beer grants the common view, "No wonder people say that management is an art. It is." (ibid.), and holds that the two descriptions do not compete: "To say that there can be a science of management is not to deny that management is an art" (MS 1967: 12). His working definition of science is broad: "systematic knowledge about the world" (MS 1967: 10). On this definition managers need not become scientists. What is required is that systematic knowledge be brought to bear on what managers do.
That knowledge offers a way of seeing the enterprise. OR gives management "a capacity to visualize the enterprise as an organic whole, to determine the structure of the underlying system that makes it what it is, and to compute in terms of the dynamics of that system" (MS 1967: 26). Understanding comes before technique: "The vital activity is the investigation of system: What makes things tick?" (ibid.). Applying a technique without that understanding is, in Beer's image, "like copying out the prescription that did Mrs. Smith so much good, and hopefully applying it to oneself" (ibid.).
Beer separates the science of a subject from skill in it: "The science of a subject is always about its very nature; it is not about virtuosity" (MS 1967: 13). Science begins with measurement but "is basically concerned with finding patterns of facts; and in particular with finding patterns that reproduce themselves", because the understanding of such patterns gives the capacity "to predict" (MS 1967: 18). Management science accordingly takes the manager's work as its object, and the shop floor only in passing: it "deploys a skill and methodology originally developed for the investigation of nature at large to do research into the manager's own job. It is only trivially concerned with the processes being managed" (MS 1967: 19–20).
From the wartime origins of OR Beer draws three lessons. Management deals with situations "about which information is essentially incomplete"; an OR team "must represent as many scientific disciplines as possible"; and strategies must exploit "the entire system within the manager's command \[...\] as a whole", in contrast to "the orthodox analytic approach, whereby the system is examined piecemeal" (MS 1967: 27). The same summary gives the reason for the third lesson: "Small advances made in particular localities of the system may actually turn out to be inimical to the health and success of the whole" (ibid.). The second lesson explains why OR "became at once, and remains to this day, an interdisciplinary activity" (MS 1967: 23); chapter 1, section 1.5 shows how Beer later preferred the word "transdisciplinary".
What OR investigates, then, is the system that produces a situation. OR "is trying to discover the nature of the underlying system that generates the particular situation under study. If we know what that system is \[...\] then we acquire predictive power" (MS 1967: 24). The manager knows his "difficulties", but these are symptoms; "the problems that need solving \[...\] are the underlying things of which difficulties are mere symptoms", and they "are locked up in the nature of the system that generates them" (MS 1967: 29). It follows that extrapolation fails whenever the proposed change alters the system. A policy for change based on "the extrapolation of the results into a future that will have been made different by that very change, is probably wrong" (MS 1967: 25).
## 27.2 Chance and variety
Chapter 2 of MS ("Chance, Risk, and Malice") introduces the quantities a calculus of decision has to handle. The first is chance. Beer attributes to managers, not to scientists, the view that science deals only with determined matters, and answers that modern science "is therefore especially designed to compute with probabilities rather than with certainties. And the scientist, like the businessman, sees his universe not so much as a collection of definite things but as a flux of uncertain interactions" (MS 1967: 32). The manager need not do the mathematics; "he ought to know what mathematics to ask for" (MS 1967: 33–34). Nor is the human element beyond such treatment: "the human being imports uncertainty into a situation just as any other kind of natural variable imports uncertainty, and the effects are no different" (MS 1967: 48).
The second quantity is variety, and with it MS makes its first explicitly cybernetic statement. As controller the manager needs "a quantified understanding, of the number of alternative combinations of events with which he has to deal" (MS 1967: 37). Beer then states the law: "According to the science of cybernetics, which deals with the topic of control in every kind of system \[...\], there is a natural law that governs the capacity of a control system to work. It says that the control must be capable of generating as much “variety” as the situation to be controlled. Now variety may be roughly defined as the possible number of distinguishable states in a situation" (MS 1967: 37). This is the Law of Requisite Variety, treated in its later formulations in chapter 2, sections 2.1–2.3, and in its DC form in chapter 25, section 25.4.
Variety grows fast. A production unit with ten processes already has "nearly ten million possibilities" (MS 1967: 38). That reverses the usual question about simple controls: against the belief that a simple control "cannot go wrong", Beer asks "whether it can go right" (MS 1967: 38). The same pair of phrases returns in chapter 6, where the governor sets the standard for a control that cannot go wrong (27.7). Seen from the other side, a situation is "just one choice out of millions of available choices", and "management becomes a task of chopping down possibilities to one actuality, rather than of building up small components to make an interacting whole" (MS 1967: 39). Management as selection among possible states, rather than assembly of parts, anticipates the later language of variety attenuation (chapter 2, section 2.4).
*Textual note:* MS defines variety "roughly" as "the possible number of distinguishable states in a situation" (MS 1967: 37); the BOTF glossary has "the total number of possible states of a system, or of an element of a system" (BOTF 1981: 403). MS adds the qualifier "distinguishable", which ties variety to what can be told apart and so to an observer — a dependence that DC makes explicit (chapter 25, section 25.3).
The chapter's examples are stocks and queues. "Stock is quite simply, and always, a shock absorber between two interacting sets of probabilities" (MS 1967: 50). Its size is a convolution of input and output distributions, and the manager's part is to state the risk he will accept (MS 1967: 52–53). Once that policy is set, a computer that reorders stock "is actually taking a decision". Beer calls this "a signpost to freedom. If we can offload rather low-level decisions of this kind to machinery by the use of science, the manager is freed to attend to more important matters" (MS 1967: 53). A related example is a reporting scheme that prints only cost variances outside statistical limits, so that "the manager gets one page of cost returns for every twenty he used to get, and the only information that has been suppressed is statistically guaranteed to be of no interest" (MS 1967: 48). The later books call both cases filtering variety before it reaches management. Beer ends the chapter by recommending the word "stochastic" to every manager: "A stochastic process is about the results of convolving probabilities—which is just what management is about, as well" (MS 1967: 58).
## 27.3 Conceptual model, scientific model, simulation
Chapter 3 of MS ("Quantified Insight") gives the model theory of *Decision and Control* (see chapter 24) in non-technical form, and it starts from the manager. His understanding "may be thought of as some kind of model of the situation out there", a "conceptual model", and the stake is practical: "bad and unprofitable decisions are the outcome of a misconception about the way the system actually works" (MS 1967: 63–64). Such a model simplifies. It maps "many-to-one"; "A good model is a homomorph" (MS 1967: 64). Beer adds: "it may well be that all models are homomorphic rather than strictly isomorphic—except one. This is the model that reality constitutes of itself" (MS 1967: 66). The remark returns as the base of the cones of resolution (27.4).
The scientist does not know the particular firm. He brings experience of other systems: "What has the scientific experience in common with the managerial experience? The answer is quite simply: system" (MS 1967: 67). His conceptual model "is a scientific analogy" (ibid.), and an analogy alone is not enough: "Unless there is such a sense, the analogy is a literary trick—not very scientific. It is no use simply stating an analogy, and just hoping that it is valid" (MS 1967: 68). Rigour comes from formal languages, since "science commands several formal languages that destroy ambiguity and equivocation": mathematics, mathematical statistics and formal logic (ibid.). The scientific model (SM) is "a rigorous statement of the conceptual model (CM)" (ibid.), and it answers how far the analogy holds. A diagram links science (S), languages (L), scientific model (SM), techniques (T) and the real situation (R) (MS 1967: 69); the same scheme in DC is discussed in chapter 24, section 24.3.
Two clarifications follow. A scientific model written in mathematics is still not a mathematical model: "The basic nature of a scientific model is *systemic*, rather than mathematical" (MS 1967: 89). And models are judged by usefulness rather than truth, because "truth and falsity are attributes of statements about the world, and not attributes of the world itself"; models are "more or less useful" (MS 1967: 72). Chapter 24, section 24.5 discusses the related distinction between models, laws and conventions.
Models are needed because experiment on the real firm is impossible, and because a model has to be tested over time, not only fitted at one moment. "The situation the manager controls is essentially viable: It goes on living" (MS 1967: 81). This is the first of several places where MS uses the word *viable*. The scientist therefore "experiments on the model instead of on the firm. If the model of the company goes bankrupt, nobody cares" (MS 1967: 82); "This procedure is known as simulation" (ibid.). Simulation also tests robustness. The manager "does not necessarily want the *most* profitable policy, if this is really rather insecure. He would prefer to trade a little of the profit for a guarantee of stability" (MS 1967: 87). The preference for stability over maximum performance anticipates the governor of MS chapter 6. Finally, a validated model "is an asset that the firm has bought"; with "learning facilities" built in, "what is happening to the firm also happens to the model" (MS 1967: 87–88). A model that runs alongside the firm and is updated by it corresponds to what the later books require of System Four (chapter 10, section 10.2).
## 27.4 An alphabet of models and the cones of resolution
Beer runs through models from acoustics, biology, cybernetics, demography, engineering, fluid dynamics and genetics, all "within my own experience of actual cases" (MS 1967: 109–110). Chapter 4 of MS uses them to show that the scientist's analogies can come from any science. The premise: "Nature is an indivisible unit; but man's way of investigating nature is compartmentalized"; the unity of nature's systemic character is being realized "in terms of a body of knowledge called General Systems Theory" (MS 1967: 91). Hence: "There is some sense in which any system can be mapped onto any other" (MS 1967: 94). DC had put the same conviction in terms of viability: "what is robust, dynamic and viable about the real-life situations which the manager handles is mirrored in the processes of nature which scientists study in all their forms" (DC 1966: 115).
Three entries bear on the later model. Under B for Biology, discussing the optimal size of a firm, Beer asks: "why should we not talk about the firm as a viable organism?" (MS 1967: 95). Under E for Engineering the market–factory loop is treated as a servomechanism. Without good control engineering "the whole system goes into oscillation"; "In engineering terms, the system is hunting" (MS 1967: 103–104). Beer presents this as a checkable prediction: "we ought to ask whether the prediction is borne out or falsified in practice" (ibid.).
Under C for Cybernetics he turns to the structure of management itself. A reliability calculation shows that a single unreliable decision cell is wrong more than half the time, while a redundant arrangement reduces the risk to about "one in a hundred million" (MS 1967: 99–100). "The new cybernetic model is a neurophysiological mapping of the way things really are. It gives a logic and a decision calculus for working out good management structures, too" (MS 1967: 100). The family-tree organisation chart, by contrast, "tells us nothing we may want to know about the way the system works" (MS 1967: 98–99). *Brain of the Firm* makes the same point: "brainlike systems are not organized into a pyramidal command structure looking like a family tree" (BOTF 1981: 232). Chapter 24, section 24.4 describes the neurophysiological model of the firm from the 1960s, and chapter 4 its later form.
The chapter ends with the question of detail. The models "preserve those structural facts that are important to the problem, but they make life easier by transforming the elements of the real-life system onto a many-to-one basis" (MS 1967: 110). How many-to-one is a choice: "We adjust the lens in order to see more or less detail as required" (MS 1967: 111). Beer calls the range of possible resolutions the *cones of resolution* (MS 1967: 113). At the apex is a model with no resolution. At the base lies "the uniquely isomorphic model: the company itself interacting with the market itself", which is of no use because "the only way to play the game at this ultimate level of modeling is to continue managing the situation as it is" (ibid.). The rule is economy: "Clearly, the most economic way to do management science is to stay as high up in the cone of resolution as one can" (ibid.). Where the scientist must go all the way down, "this is where the old-fashioned empirical research really begins" (MS 1967: 115). The result is "a hierarchy of models rather than one model. His first model is coarse, his second finer, and so on down. The finer models deal with parts of systems, and lock into coarser and more complete models higher up the cone" (MS 1967: 115).
Models nested within models, each describing the whole at its own level of detail, bear a family resemblance to the later principle of recursion (chapter 6). The texts do not say that Beer derived one from the other, and the cones of resolution concern the scientist's models, not the structure of the firm.
## 27.5 A worked case: interactions, self-organization, identity
Chapter 5 of MS ("It Works") shows the hierarchy of models in two cases from practice. Beer first warns against techniques used alone. Mathematical programming "is best thought of as one of the techniques by which systemic models may be quantified", and the usefulness of an OR team "lies in its ability to make a model of the whole (relevant) enterprise" (MS 1967: 122).
In the first case the conceptual model "was drawn from ecology, because this is the science that considers an organism interacting with its environment" (MS 1967: 123). It had two boxes, company and environment, in a closed loop, with nothing yet specified inside them. Even so, conclusions about stability could be drawn, because they follow from the relationships: "Its general behavior is determined more by interactions than by the things that interact" (MS 1967: 126). The same priority of relations over components runs through Beer's later accounts of holism (chapter 26).
The next model had to locate the decisions, and Beer found that the system makes some of them itself: "sometimes the system itself effectively makes the choice"; "we are talking about the *self-organizing* properties of systems" (MS 1967: 128). "Management relies more heavily than it often realizes on the system's intrinsic ability to do the right thing" (ibid.). The general principle is a variety argument: "If the managerial systems we create were not basically self-regulating, we should never be able to manage them; they would generate too much complexity for us to cope with" (MS 1967: 130). Self-regulation makes the managed system manageable at all, and the same argument later grounds the primacy and autonomy of System One (chapter 7, section 7.2). The servo model showed "by default, what was *not* self-regulating: That is to say, where managerial decisions had to be taken, on what evidence, and to what effect" (ibid.). Beer adds: "Few businesses know these truths about themselves" (ibid.). The chain ran from ecology through general systems theory and servomechanics to a mathematical program, so that the final computation "is actually *about* the real-life situation" (MS 1967: 131).
The second case, siting a factory, turns on identity. The distribution network is one of "those systems that have a total identity that means more than the sum of the parts"; "it is a *machine for distributing*" (MS 1967: 132). "Its integrity as a system depends upon these particularities. Therefore, if we seek arbitrarily to change the location of a major factory, we destroy that machine's identity and integrity" (MS 1967: 133). In the later books, identity as something a system has and can lose becomes part of the definition of viability (chapter 5, section 5.2).
## 27.6 Control as the continuous function
"Much more important, because it is continuous and sustained, is the manager's function of control" (MS 1967: 141). With this sentence chapter 6 of MS, "The Viable Governor", turns from policy and decision to the third function, and the book comes closest to cybernetics. Beer defines being in control by behaviour after a disturbance. A system in control returns to its pattern, and "The best word for “being in control” in this sense is *stability*" (MS 1967: 143). He warns against an indirect criterion. Systems in control usually look smooth, so people may take smoothness as the evidence; if so, "they may come to regard smoothness instead of in-control-ness as the object of the exercise, *and this will make them resistant to change*" (MS 1967: 144). The appearance of order can become a goal in itself and obstruct adaptation, a theme that returns in 27.11.
From this he introduces the discipline: "if cybernetics is the science of control, and if management might be described as the profession of control, there ought to be a topic called *management cybernetics*—and indeed there is" (MS 1967: 144; compare DC 1966: 254). The formula comes from DC, which states it twice. The introductory overview reads "Now if cybernetics is the science of control, management is the profession of control", adds that this holds in a certain type of system, and concludes that one may recognise "the subject of management cybernetics" (DC 1966: 239). The chapter on cybernetics draws the consequence for OR: "For cybernetics is the science of control; and management is the profession of control. It follows that models drawn by the OR scientist from cybernetics have a direct bearing, an immediacy, which models drawn from other sciences lack" (DC 1966: 254). In MS the formula introduces a particular conception of control, set out in the next sections. Chapter 1, sections 1.1 and 1.4, follow the definition of cybernetics through the later books.
## 27.7 Implicit control
The model is the Watt governor, in which "Controllability is *implicit* in the operation of the machine" (MS 1967: 145). Beer sets it against a prison governor who controls by periodic inspection, where the lags between inspections defeat the control (MS 1967: 145–146). The steam governor "is an implicit controller: Variables are brought back into control *in the act of and by the act of* going out of control" (MS 1967: 147). There is no separate act of checking and correcting. Implicit controllers "depend for their success on two vital tricks": "continuous and automatic comparison of some behavioral characteristic of the system against a standard" and "a continuous and automatic feedback of corrective action" (ibid.).
The same principle appears as "intrinsic control" in BOTF (1981: 86) and HOE (1979: 91); see chapter 2, section 2.7. HOE ties it explicitly to requisite variety: "This regulatory process of INTRINSIC CONTROL sees to it that Ashby's Law is automatically obeyed" (HOE 1979: 91). BOTF applies it to the firm: "From the standpoint of senior management, that is to say, there are intrinsic controllers at work which do not, or do not normally, demand attention" (BOTF 1981: 86). It then complains that many senior managers "not only attend to these implicit controllers but interfere with them and stop them working" (ibid.), using the word that MS had used.
Beer applies the governor to management practice. "There is a difference between a self-regulating system that one confidently expects to go right, and a self-regulating system that cannot go wrong. Management, I fear, almost always uses the first. It ought to be the second" (MS 1967: 147). "Management by exception" is "a way of trying to express the need for implicit control", "But this is not enough" (ibid.); in practice "it becomes impossible to detect what really is an exception, until it is too late" (MS 1967: 149). The reason is requisite variety, invoked for the second time in the book: "the variety of the controller must be at least as great as the variety of the situation to be controlled, and this requirement is likely to be falsified" (MS 1967: 149). The remedy is to change what is measured. Beer proposes "an output of our system that we expect to remain steady", such as the ratio of planned to actual time, which "reduces variety straight away" (MS 1967: 150). Choosing a measure that should be stable and controlling deviations in it is the move the later books make with their indices of performance (chapter 16, section 16.5).
## 27.8 Ecology: viable governors
The steam governor controls one variable in a designed and fully known machine. Most managerial systems are not like that. For systems whose variables cannot all be identified, Beer turns to ecology: "The fact is that living systems—interacting animal populations—are intrinsically controlled. They are *viable* governors" (MS 1967: 151). The chapter takes its title from this sentence.
Management usually controls high variety "by dividing them into subsystems" with "sufficient sub-controllers to absorb the variety"; this "is sound practice in cybernetic terms, but it is not the only nor the most economical approach to the job" (ibid.). An ecological system has no such apparatus. "There are no managers, no controllers, no sub-controllers, no bureaucracy, and no paper work"; there is "a vast network of interacting governors", and "no unique feedback, no unsupported control loop, on which the whole system depends" (MS 1967: 153). Beer concludes that "just because of this, the whole business is an excellent model of business itself" (ibid.). Stability from a network of mutually regulating parts, with no single loop on which everything depends, connects with the redundancy argument of 27.4. It also connects with the later treatment of System One, whose parts regulate one another before any metasystem intervenes (chapter 7, section 7.4, and chapter 8).
## 27.9 Homeostasis and ultrastability
Beer credits Ross Ashby with the discovery of these self-regulating properties and names the kind of self-control homeostasis (MS 1967: 153). A homeostat "holds the critical variables of the system stable within *physiological* limits", and these limits "are not arbitrarily imposed from outside, they are generated from within" (ibid.). Here the governor's standard has moved inside the system. The example is body temperature, held at 37°C "because the organism itself recognizes its own optimal working temperature" (MS 1967: 154). The management interest is immediate: "if it is possible to have a self-regulating system that implicitly arranges its own stability, then this is of the keenest management interest" (ibid.).
The second property is ultrastability. A servomechanism copes with known and nominated inputs, but big managerial systems "are indeed shaken by all sorts of disturbances that we may not have thought about when we arranged the set of management controls" (MS 1967: 154). Ultrastability is "a capacity in a system to regain equilibrium after any kind of perturbation, including kinds the designer did not have in mind" (MS 1967: 155). BOTF also credits the concept to Ashby and places it in physiology: it "was devised by Ross Ashby specifically to explain, also in mathematical terms, the nature of physiological homeostasis" (BOTF 1981: 144).
Beer limits the claim: "There is, in short, a threshold of disturbances beyond which the system is denatured, if not actually annihilated" (MS 1967: 156). The mechanism draws its power from not needing to know the causes of disturbance. It "recognizes the results of disturbance rather than the nature of disturbance, and takes action to avoid unpleasant consequences without attempting to resist their unknown causes" (MS 1967: 158). In the later books this becomes the principle of the black box: regulation works on observed behaviour, without a full account of the mechanism (chapter 2, section 2.5).
*Textual note:* The definition of ultrastability in MS ("any kind of perturbation, including kinds the designer did not have in mind", MS 1967: 155) and in the BOTF glossary ("perturbation by unknown or unanalysed forces", BOTF 1981: 403) agree in substance but differ in wording.
## 27.10 A new model: company and market
Beer then draws company and market as two boxes of dots: "Each dot represents a possible state of the company or the market" (MS 1967: 157). The dots stand for each system's variety, and a circle encloses the preferred states. "Homeostatic equilibrium is reached when what is happening in each subsystem results in the representative point of the other system being chosen from its preferred set of states" (ibid.). Where the Watt governor held one machine to a standard, this model holds two coupled systems within each other's preferred states. Beer notes that the equilibrium is "precisely the criterion of “smoothness” that we noticed much earlier" (MS 1967: 158), the smoothness that, taken as an end, breeds resistance to change (27.6). If one subsystem is pushed out of its circle, the whole oscillates "until both points, and not just one of them, have returned to safety"; Beer reports that "Ashby called this part of the mechanism “self-vetoing.”" (MS 1967: 158).
BOTF uses the same construction and the same picture. When two homeostatically regulated systems are coupled, "a meta-control operation may be envisaged. It works as a self-vetoing system, and is depicted in Figure 26" (BOTF 1981: 145); the caption reads: "Each dot represents a total configuration of the system, dots contained in circles represent satisfactory states" (BOTF 1981: 146). HOE uses the circle for a system's stable range, "within physiological limits indicated by the circle surrounding the dot" (HOE 1979: 381). The model of coupled homeostats with preferred states is thus present in 1967 and recurs in both later books. Its best-known later use is the homeostat between Systems Three and Four (chapter 11, section 11.1).
The real case has more parts: "the cybernetic models we use are such multiple, polystable homeostats", with "little homeostatic subsystems in every one of the boxes drawn, and \[...\] homeostatic supersystems that link the boxes together" (MS 1967: 158–159). This describes a layered arrangement of systems within systems; it is not yet the recursive five-system model.
Beer limits what the method claims: "the management scientist does not claim to make shattering new discoveries about the nature of business. He knows very well that the management understands its job" (MS 1967: 159). The aim is "to provide a rigorous account of what the manager knows" (ibid.). Such models investigate "stability, ultrastability, and the system's physiological limits", and they do not forecast in detail: "We have to face the fact that we cannot predict the future in detail. But our rigorous analysis can show that some loop of the control system is missing, that some of the laws known to govern systems of this kind are being disobeyed, or that some subsystems are placing undue stresses on others" (MS 1967: 162). Looking for missing loops, broken laws and overstressed subsystems is diagnosis. It resembles the procedure that *Diagnosing the System for Organizations* later sets out (chapters 14 and 15).
## 27.11 The education of systems
"We were surely right to think of viable governors as seeking, first and foremost, stability. An unstable organism, after all, is likely to go out of existence, or to go mad; our very first task, as men or as companies, is to survive" (MS 1967: 162). Stability is the first requirement, and Beer does not leave it as the only one. A steam governor returns to its standard as often as it is disturbed; viable systems also learn. "Learning and adaptation, growth and evolution, are really implicit in the concept of survival, even though its precondition may be a static-seeming ultrastability" (MS 1967: 163). Learning is defined operationally: "the time taken to reach a stable response to a given stimulus shortens as the stimulus is repeated"; and "Individuals are certainly learning; much more importantly, the system as a whole is learning too" (ibid.). Adaptation to real change "is really a special kind of learning, which will involve some restructuring of the system itself" (ibid.). The later books assign adaptation through structural change to the metasystem, in particular to System Four (chapter 10).
Beer ends the chapter by separating management cybernetics from OR: "The cybernetician is seeking better control structures for the firm" (MS 1967: 166). In MS, OR compares alternative courses of action and management cybernetics looks for control structures. The later books are almost entirely about the second.
## 27.12 Automation, structure and purpose
Chapter 7 of MS ("Automation and Such") applies the argument to computers. Beer denies that management science is "really a question of making computer applications" and warns against "the mania which declares that science without computers is not science" (MS 1967: 169–170). He describes industrial development with a physiological model. Mechanization supplied energy; with advanced mechanization "the body acquired nervous control up to the level of reflex" (MS 1967: 170, 173). Automation "adds to mechanization the computational abilities of a brain", and "like a brain, automation can fulfill this task only if it is provided with criteria of success"; its mark is "automatic decision-taking capacity" (MS 1967: 173). A further phase, in which a company could formulate its own policy, "has already been called *cybernation*" (MS 1967: 174). Beer reports the term rather than coining it. On my reading, the sequence of body, reflex, brain and a level that sets the criteria of decision is related to the levels of the nervous system that BOTF maps onto the five systems (chapter 4, section 4.2). MS, however, uses it for historical phases of industry, not for the structure of a firm.
Why does this not happen? Beer's answer is structural. Departments exist because "the limitations of the human hand, eye, and brain commit us to departmental activity"; the firm's anatomy "is a structure frozen out of time" (MS 1967: 175–176). "So we reach a point in understanding where we see that the way everything is structured really derives from a *human* incapacity to be structured differently" (MS 1967: 176). To automate the existing departments is to "enshrine in steel, glass, and semiconductors those very limitations of hand, eye, and brain that the computer was invented precisely to transcend" (MS 1967: 177). Automation of this kind even "reinforces the walls that separate one department from another" (MS 1967: 181).
The order of rethinking is purposes, policies, structures, problems and only then processes (MS 1967: 180, 182). For "anyone who has followed the general arguments of this book on the matter of the indivisibility of viable systems", automation offers the chance "to recreate the integral management that the firm enjoyed a century ago"; otherwise "we shall get what might be called rational but trivial automation" (MS 1967: 182). That information technology should serve a redesigned structure rather than freeze the old one is also central to Beer's account of Cybersyn (chapter 16). Beer applies the same criticism to banking and transport (MS 1967: 185–187) and to government, where "Examination of purposes is restricted to the elevated discussion of liberal principles; it is not undertaken in the practical context of affairs" (MS 1967: 187). He rejects the charge that he is substituting one faith for another: "Science is not a magic word, and scientists are not demigods" (MS 1967: 188). The book closes: "This book has argued that huge, not meager, savings and improvements are made possible by the integral control of integral systems" (MS 1967: 188).
## 27.13 Place in the development of the VSM
MS contains no five-system model: no Systems One to Five, no metasystem in the later sense, no algedonic channel, no recursive system theorem. In 1967 it does contain several elements that the later books build on:
- management cybernetics as the science of control applied to management (MS 1967: 144; DC 1966: 239, 254);
- the requisite-variety requirement and a definition of variety (MS 1967: 37, 149);
- control as a property of the system rather than an act of inspection (MS 1967: 145–147);
- homeostasis within internally generated limits (MS 1967: 153);
- ultrastability (MS 1967: 155);
- the company–environment interaction as coupled homeostats (MS 1967: 157–159);
- the hierarchy of models and the cones of resolution (MS 1967: 113–115);
- self-organizing decision elements (MS 1967: 128–130);
- identity and integrity of a system (MS 1967: 132–133);
- and the word *viable*: for the situation the manager controls (MS 1967: 81), for the firm as organism (MS 1967: 95), for self-regulating living systems (MS 1967: 151) and for "the indivisibility of viable systems" (MS 1967: 182).
Some of these reappear almost unchanged: the self-vetoing homeostat with dots and circles (BOTF 1981: 145–146; HOE 1979: 381) and intrinsic control (BOTF 1981: 86; HOE 1979: 91). Others change function. The cones of resolution describe the scientist's models, whereas recursion describes the organisation; the physiological sequence of MS chapter 7 describes the history of industry, whereas the nervous-system model of BOTF describes the firm.
The list records passages. It does not claim that Beer himself described MS as a stage of the VSM. What the texts support is narrower. In 1967 Beer already described the firm as a viable system held stable by a network of intrinsic, homeostatic controls. He had already separated what regulates itself from where management must decide, and he had already stated that "The cybernetician is seeking better control structures for the firm" (MS 1967: 166).

## References

- **DC** — Beer, Stafford (1966). *Decision and Control: The Meaning of Operational Research and Management Cybernetics*. Chichester: John Wiley & Sons. Cited from the Stafford Beer Classic Library reprint, 1994.
- **MS** — Beer, Stafford (1967). *Management Science: The Business Use of Operations Research*. Garden City, NY: Doubleday, in association with Aldus Books, London.
- **HOE** — Beer, Stafford (1979). *The Heart of Enterprise*. Chichester: John Wiley & Sons.
- **BOTF** — Beer, Stafford (1981). *Brain of the Firm*. 2nd ed. Chichester: John Wiley & Sons. (1st ed. London: Allen Lane, 1972.)