On organizations as brains

Espen Andersen, 1992

This paper summarizes Gareth Morgan's (1986) view of the brain as a metaphor for organizational structure and behavior. It then discusses some important developments in cognitive science since Morgan's book, and tries to show how these developments may affect the metaphor and what we can learn about organizations from it.


In his book Images of Organizations (1986), Gareth Morgan provides an intriguing introduction to organizational theory, dividing the strands of literature into different kinds of metaphors: organizations as machines, organisms, culture, political systems, etc. The metaphor of the organization as a brain is in the third chapter, where Morgan discusses three major strands of literature contributing to this metaphor: literature on information processing (Galbraith, 1977; Simon, 1945), on cybernethics and self-correcting systems (Wiener, 1961) and on organizational learning (Argyris & Schön, 1978). He compares organizations to holograms1, and postulates four principles to create a "hologram" organization2: He outlines the strong points of the brain metaphor as but also lists two serious weaknesses: Morgan uses the brain metaphor externally, in the sense that he is not concerned with drawing comparisons between how the brain actually works and organizations, but rather uses the metaphor to anthropomorphize organizations based on characteristics of the brain as observed from the outside. The brain metaphor provides a way to approach organizational learning the way you would approach individual learning: by building trust in the teacher, enthuse commitments to learning goals and provide availability of tools.

 Since Morgan wrote his book, our understanding of how the brain works has changed, mainly through the increased scope and popularity of cognitive science. The strides made in cognitive science have had important implications for both scientific and popular views of information processing systems. What consequences will they have for the use of the brain as a metaphor for organizations?

Popular cognitive science

Cognitive science is a relatively new scientific discipline, concerned with how the brain works. It takes as its reference discipline neurology, cognitive psychology, computer science, information theory, philosophy and almost anything else that can provide a useful method or metaphor. Much of what is written (or has later been classified as belonging) under the cognitive science umbrella has been philosophical and somewhat popular in nature, like Hofstadter's Gödel, Escher, Bach (1979), which describe the difficulties in understanding systems because we tend to look at the wrong levels we try to understand where in an ant-hill the knowledge about building ant-hills resides by looking at individual ants, for example. More stringent approaches have come from the cognitive psychologists, like Card, Moran and Nevell (1983) which sees the brain as a system with perception, cognition and motoric subsystems: Perception is done in parallel, cognition serially.

 In 1986, two important books came out, both which would influence cognitive science immensely. One was Marvin Minsky's Society of Mind, which gained huge popularity and helped establish popular support for the theory of the brain as a system put together of many small, relatively simple systems. The other one was Rumelhart and McClelland's (1986) Parallel Distributed Processing: Explorations in the Microstructure of Cognition, which provided a formal theory for how the processing was done: by having many small processors connected together in immense networks, each processor capable of reaching everybody else. Connectionism, the term for these theories, has had a tremendous impact on AI, computer science and cognitive psychology, largely because, as Dennett puts it, it "blazes the first remotely plausible trails of unification in the huge terra incognita lying between the mind sciences and the brain sciences" (1991 p. 269).

 But connectionism, while providing a theory for how the brain works, does not provide a theory for how the process is managed: how the brain chooses to do what it does, that is, how it becomes conscious. The common thinking about consciousness is that it is "thoughts about thoughts": a stance which presupposes somebody or something that thinks the thoughts about the thoughts (leading to infinite regressions, "thoughts about thoughts about thoughts"). Daniel C. Dennett (1991), based on these theories and a number of others, tries to extend the "society of mind" theories to include how the whole brain is governed: how consciousness develops, is invoked and operates. His central thesis is that of "Pandemonium Theory":

There is no single, definitive "stream of conscious- ness," because there is no Headquarters, no Cartesian Theater where "it all comes together" for the perusal of a Central Meaner. Instead of such a single stream (however wide), there are multiple channels in which specialist circuits try, in parallel pandemoniums, to do their various things, creating Multiple Drafts as they go. Most of these fragmentary drafts of "narrative" play short-lived roles in the modulation of current activity but some get promoted to further functional roles, in swift succession, by the activity of a virtual machine in the brain. The seriality of this machine (its "von Neumannesque" character) is not a "hard-wired" design feature, but rather the upshot of a succession of coalitions of these specialists.

 The basic specialists are part of our animal inheritage. They were not developed to perform peculiarly human actions, such as reading and writing, but ducking, predator-avoiding, face-recognizing, grasping, throwing, berry-picking, and other essential tasks. They are often opportunistically enlisted in new roles, for which their native talents more or less suit them. The result is not bedlam only because the trends that are imposed on all this activity are themselves the product of design. But it is augmented, and sometimes even overwhelmed in importance, by microhabits of thought that are developed in the individual, partly idiosyncratic results of self-exploration and partly the predesigned gifts of culture. Thousands of memes, mostly borne by language, but also by wordless "images" and other data structures, take up residence in an individual brain, shaping its tendencies and thereby turning it into a mind.

 (1991, pp. 253-4)

Equipped with this rather cursory overview of some theories in cognitive science, let's turn to a look at how the workings of organizations can be shown to follow a similar pattern.

Consequences for the organizational metaphor

When looking at organizational processes, viewing the organization as an information-processing system, we can divide what is going on into two main activities: The retrieval and dissemination happen when the organization, or some part of it, is trying to solve a small or large problem, here referred to as organizational decision processes, although the decisions may not be recognized as such "in process". The storing of information includes the selection of what is to be stored, where it is to be stored, and how it is to be retrieved. We may refer to this as organizational memory. The control of these processes, to the degree that they can be controlled, I will refer to as organizational governance.

Organizational memory

Morgan's concept of the organization as a brain is based on the way organizations store information; that is, how organizational memory functions. As found by animal experiments, large parts of the brain can be removed without significantly impeding the functions of the organism. Organizations may exhibit the same feature: a small Norwegian shipping company lost half of its employees, including many managers, in a charter plane crash. After the initial shock had subsided, the organization was able to function very much like before. The knowledge of the members of the organization that died was duplicated in the remaining half, not in the sense that there were two of every kind, but in the sense that by pooling their knowledge and creating new organizational structures, the employees were able to reconstruct the functions performed by the people who perished.

The ability to survive the disappearance of working parts of the system does not strike us as particularly difficult in an organizational setting: after all, humans are adaptable and can work harder under difficult situations than under normal conditions. We can easily imagine how each person in the shipping company behaved, both in their own work and in relations to others, to make the organization survive. This is a case where the organization can teach the cognitive scientists something: It is more difficult to understand that our brain may work in exactly the same way: that knowledge (including knowledge of internal working procedure) may be stored and accessed like an organization stores and accesses knowledge.

Walsh and Ungson (1991) posit that organizations have memories, and that they consist of five "bins" or retention facilities: individuals, culture (language, shared frameworks, stories), transformations (how things change, procedures), struc- tures (institutionalization) and ecology (primarily physical settings). Added to this comes external archives, such as industry indexes, publications, etc. Interestingly enough (particularly for information systems designers) internal archives are not considered a very important part of organizational memory, although this might be thought of as just an implementation issue of the five bins above.

The distributed character of organizational memory has important consequences for how things are done: since what the organization previously has learned plays an important role in what it will do in the future, actions to change an organization must take into account how knowledge is stored and retrieved. In a bureaucratic organization, information is collected by the lower ranks, collated and presented by middle management for the top, who decides base on it and gives orders back. The drive towards autonomous teams and decentralization in the face of an increasingly changing environment is in my view a tacit acknowledgment that organizational memory is spread around in the organization, and that the best way to access it is through moving the decisions to it, rather than the other way.

Organizational decision processes

The brain is an incredibly powerful information processing system: although each neuron is not a particularly powerful processor in itself, in fct not even close to the speed and processing power of a modern microprocessor, the vast network of all the neurons in a brain nevertheless process information in fantastic quantities and speeds, seemingly effortlessly. Minsky's Society of agents and Dennett's pandemonium theory seems chaotic, but the brain's performance of this process beats all information processing system likely to be around at least for this century.

The reason the brain can process information so quickly is its exploitation of massive parallellism: instead of, like most computers, having one single, albeit extremely fast, processor chugging through everything, the work is divided up and performed by a number of specialists: parts of the brain, not specifiable (though activities are localizeable to a certain extent) but interrelated, mingled together. Things are figured out by testing what Dennett calls Multiple Drafts: possible scenarios, which are tested for credibility and then chosen or thrown away based on how much support they gain.

The whole process is somewhat similar to what Quinn (1980) calls logical incrementalism: the gradual process through which organizational strategy is formed. Top management, in Quinn's view, will do wisely in just letting strategy percolate up through the organization, first because it is likely to be better, second because it will not be resisted to the extent that dictums from above might be. The problem is that the process takes so long: direct orders at least let everyone know what to do immediately.

Or do they? John Akers has been trying to turn IBM around for years, but the incredibly complex, giant company just isn't doing it. The reason might be that most of the employees have not understood the severity of the need for change (IBM will always pull through), or that they do not understand the orders (and they are very abstract). The more likely reason is, in my view, that the top-down process used by Akers does not sit well with what he is trying to do: institute behavioral changes. Since IBM does not have a process for directly doing this, it has to take the circuitous route of structure: by providing a structural solution to a behavioral problem (Argyris, 1990), IBM's top management preserves the hierarchical process they have always relied on.

Changing IBM (or any large organization) is a huge task. Changing the way work is performed at an individual level is a huge task, too, but is doable by that very individual, given the right tools. In other words, this is a task that is well suited to parallell processing--if we could just figure out a way to divide up the job and disseminate the tools used to do it. Parallell processing has, in the computer field, showed that networked individual workstations can together provide computing power equal to Cray supercomputers at a fraction of the cost4. The problem, however, lies not so much in physically tying together the processors, as in dividing up the problem and putting together the answers afterwards5.

I would submit that what is now happening (or rather, not happening) at IBM would be much easier understood in terms of a brain metaphor than any other organizational image. The employees at IBM, who have been working in a stable (yearly reshufflings aside) environment for many years, just cannot take messages that the organization, and themselves, have to change dramatically. Realizations of this kind, I believe, have to come from personal understanding: and understanding that must work its way up from below. Ford Motor company has been partially successful in turning around, mainly through a number of grassroot-like movements supported by management (Pelofsky, et al., 1989). This breaks habits in the organization because it breaks habits in individuals: organizational processes change not because they are externally (that is, by management) ordered to change, but because the individuals performing them change how they do things. The role of management changes from being the order-givers to being the "keepers of the flame": trying to spot what is good and transfer it to other parts of the organization. This is coherent with Dennett's view of the processes in the brain as "...augmented [...] by microhabits of thought that are developed in the individual.." (see above). By changing the organizational "microhabits", we can change organizations.


Although hierarchies seem to have been a feature of human life and society at least since the Cro-Magnons, the functional hierarchy in organizations was introduced by Max Weber in the late nineteenth century (Weber, 1947). Based on the public organizations in Bismarck's Germany, his bureaucratic model of organization had the hierarchy of control as its central tenet: the top of an organization decides what to do, and the units below the top execute, in increasingly fine-cut specialization as the distance from the top increases. The head of an organization was the only one who was free to do what he wanted: the ability of the rest of the organization to follow his lead quickly and without asking questions was the power and economical rationale for the bureaucratic form.

Although many authors, have pointed out that this maneuverability is impeded by a number of features of the bureaucratic form itself (Crozier, 1967; Merton, 1940), the hierarchical organizational form remains the dominant today: even futuristic articles rarely question the role of the central, direction-finding function. Instead they are more concerned with the interface between what to do and how to do it: the reduction of layers between the top and the people actually doing the job (Applegate, Cash & Mills, 1988; Drucker, 1988; Leavitt & Whisler, 1958). Indeed, I've yet to find any article about organizational structure, directed at practitioners or academics, that does not start its description of an organizational form with the singling out of a small group which are to lead the organization.

It may seem that the idea of the masses led by the few is inherent in human behavior: there has to be a central leadership. The various political theories do not question the necessity of the leadership, but are more concerned with the process through which it is established and how the positive attributes of it (representativeness, justice, effectiveness) are maintained. Chosen, inherited, appointed or God-given, there has to be centralized decision-making, at least for organizations over a certain size. From a system theoretic perspective, the hierarchy increases robustness of a system because it ensures that environmental influence does not destroy the entire system, but merely the part currently worked on (Simon, 1962). This gives the hierarchy, or at least the modular quality of it, an evolutionary advantage leading to its dominance in all systems found in nature and society. The bureaucracy is robust in the sense that offices are, at least in theory, independent of who occupies them.

But a distributed, flexible system like the brain is just as, maybe even more, stable than a functional hierarchy. Moreover, it is the status quo in many, if not most, organizations, due to increasing professionalism and a change from physical products to information as the outcome of an increasing proportion of work performed. Organizational processes may seem to be run according to Weber's principles, but the location of knowledge and the need for individual behavioral change (and the difficulty in measuring and inititating this change centrally) ensures that decision processes are distributed rather than centrally controlled.

Then what about governance? The top management does not function as the locus of knowledge, or the only policy decision-maker. Then what do they do?

I would venture that the main role of top management in a modern organization is to be the embodiment of the organization's "self": a cognitive crutch for everybody in and around the organization. This is recognized by Mintzberg (1973) whose list one of managerial roles includes "figurehead", "spokesman" and "disturbance handler". Top managers, especially the CEO, personify the organization, much as the concept of soul or a "cell or group of cells of such anatomical and functional preeminence as to appear to be the keystone or center of gravity of the whole system" (William James, 1890, as cited in Dennett (1991), p. 101). Even though we know that it is not true, we need the concept of a focal point to anchor communication and conceptualization to and about the organization. Of course, all organizations employing the same mental model, there must a top: The idea of communicating with an organization without a center (or top) would seem suspect: there is nothing reassuring about a cloud, even though it may be a top-performing cloud.

Dennett's pandemonium theory provides an answer to the question: what is consciousness, and consequently, what is self: "the organization of information that has structured your body's control system", or, more provocatively, "the program that runs on your brain's computer" (1991, p. 430). Transferring this to an organizational setting, the organization is what it is doing, not the entities performing the work. The notion that the executive office personifies the organization implies that it knows everything about it. This is clearly not so, even though the top of an organization, due to its say in matters of recruitment and rewards, has disproportionate influence. In an increasing number of companies, the most important people are not those who administer, but those who are good at what the organization is doing: those that embody the organization in themselves.

It would be tempting to draw this anthropomorphism further, and raise the concept of organizational consciousness. I think this is going a bit too far, since no present organization is even close to the complexity of even simple brains like those of insects. We do not know the mechanics behind consciousness in the brain--merely that the brain produces it, through processes infinitely more complex than even the most complex undertakings in organizational learning. It could be argued that even if organizations had consciousness, we would not be able to see it, since we would only see parts of it. Moreover, the communication of organizational consciousness would be in different time frames (much like A Conversation with Einstein's Brain (Hofstadter & Dennett, 1982) or like starfish, which to us seems just to be lying around, but really have a very active social life in extremely slow motion). On the other hand, organizations show signs of culture and personalities (ask anyone who ever has dealt with IBM versus Apple Computers). They may enact their environment (Weick, 1979) and be aware or unaware of problems in it.

I think the concept of organizational consciousness (or, rather, of organizational "self") is of value, because it offers a role-set for managers, professionals and other participants in an organization, a way out of the concept of hierarchy, a robust solution to the riddle of how to effectively run a post-bureaucratic, non-hierarchical organization. A realization that the chief role of top management is to embody the organizational self rather than having the organizational self embody themselves, will help us because such an inclination may provide a viable path in the change-over to the "new organization" Peter Drucker (1988) foresees.


The features of the brain, outlined above, in my opinion seem to characterize much of what goes on in organizations. Like the brain, organizations have a structure imposed on it, in organizations in the form of organizational charts, in the brain in the form of "maps" outlining which part of the brain handles what. In both instances there are acknowledged discrepancies between the map and the terrain6: we know that things are not like on the organizational chart (although some things are), but the chart nevertheless is produced, ostensibly as a communication device, in reality because it perpetuates the idea that organizations are run top-down. An increasingly changing environment shows the vast untapped potential of the organization, untapped because we do not know the terrain, that is, the informal organization, and this worries us.

Metaphors are important because they are the tools of thought (Dennett, 1991 p. 455). The metaphor of the organization as a brain has previously been used to illustrate the concepts of autonomous groups and organizational learning. It is becoming increasingly viable because of the advances in cognitive sciences, which now provide a theory not only of the externally observable behavior, but also of the mechanics, of an organization as a brain.



  1. One feature of holograms is that, if shattered, each single piece can be used to reconstruct the whole picture.
  2. I base my interpretation of Morgan's book on a Norwegian translation, which means that citations from the book not necessarily are identical to Morgan's text.
  3. See also Simmons (1990).
  4. See, for instance, "David Gelernter's Romance with Linda", New York Times, January 19, 1992, for a somewhat fanciful explanation.
  5. Interestingly enough, IBM might be one of the few companies in the world which have the information technology infrastructure around to do just that. Every IBM employee has a mainframe account with a unique name, and is tied into a vast network of mainframe computers, called VNET, through which he or she can communicate with any other IBM employee. In theory, all these employees (385,000 at the last count) should be able to participate in the transformation of IBM through communicating over these networks--discussing ideas and putting them into practise. In practice, however,this is not likely to happen, simply because the software used to communicate with is not designed for discussions in groups (that is, computer conferencing) but rather for sending off memos to a distribution list: to have a discussion, all discussants must have a copy of the list, and send a physical (albeit electronic) copy to everybody else.
  6. See Dalton (1959) for an early analysis of "official" and "real" organizational maps.

Copyright © 1992 Espen Andersen.
This page at http://www.espen.com/papers/orgbrain.htm.
Contact information at http://www.espen.com/contact.html.