For years, organization theory has been pulled in opposite directions by the implicit themata of economics and of biology (Holton, 1973). Economics, on the one hand, has been Platonic and Newtonian in its orientation (Mirowski, 1989). It evolved with a focus on stable equilibria and, for that reason, found it difficult to handle discontinuous change—until recently, comparative statics was as near as it got. Economics also tried, in a reductionist fashion, to minimize the part played by context in its explanatory schemes. Biology, on the other hand, has been more Aristotelian and context oriented. It has also been more willing to admit of irreversible time-related processes and far-from-equilibrium phenomena. All biologists have to come to terms with irreversible change, if only in their own lives!

The term “organization” is common to both biology and the social sciences. Within the latter, the term is of particular interest to the fields of management and economics. Indeed, the organization of firms and the organization of industries can be said to form a large part of the core of these fields of study. But do biologists, management scholars, and economists refer to the same underlying phenomena when they use the term “organization”? Is there a common set of concepts that underpins the use of the term in these different disciplines?

The term “organization” is clearly not the exclusive property of organizational scientists. Biologists, on etymological grounds alone, have a greater claim to it. Indeed, students of social and economic organization have been much influenced by biological thinking, as witnessed, for example, by work on the population ecology of organizations (Hannan & Freeman, 1989; Carroll, 1987; Aldrich, 1979), on product and organizational lifecycles, on organizational growth and development, etc. In each case the borrowing has been partial, often undisciplined, and sometimes arbitrary.

Until the 1980s, the closest that organization theory came to thinking in rigorous biological terms was through general systems theory and cybernetics (Von Bertalanffy, 1976; Katz & Kahn, 1978; Ashby, 1954; Wiener, 1962). But, as Paul Cilliers observes, general systems theory and cybernetics are ahistorical in character, i.e., they are reversible and have no memory (Cilliers, 1998). To that extent, they represent a mechanistic tradition within biology rather than an organismic approach to organizational issues. For biology in its early days, under the spell of Descartes’ mechanical philosophy, was almost exclusively mechanistic—think, for example, of Harvey’s pump model of the heart or of de la Mettrie’s homme-machine.

The question we wish to address in this article is whether, when they talk of organization, the biological and managerial disciplines are holding on to different parts of the same animal or on to different animals that merely show a passing resemblance to each other. Certainly, three decades ago, organization theorists believed that general systems theory offered them a theory of organization that they intimately shared with biologists. But systems theory à la 1960s (which was never part of the biological mainstream) has been superseded by the complex adaptive systems (CAS) paradigm, and organization theorists have not yet fully adopted this new thinking. Their implicit models still owe more to the 1960s than to the 1980s or the 1990s. Biologists, for their part, have in general moved on, accepting complex systems thinking in disciplines as different as ecology and endocrinology. Does the concept of complex adaptive systems move the two disciplines closer to each other? Does it foreshadow a universal theory of organization?

Both social science and biological concepts of organization have to deal with the problem of agency and hence with the problem of knowledge. Yet, whereas in the social sciences knowledge has been framed primarily as a cognitive issue, in biology knowledge shows up more as the driver of purposeful behavior, whether or not cognition as commonly understood is involved. In economics, for example, knowledge is a cognitive object. In contrast, for biology, knowledge is a disposition or capacity to respond. As Popper put it, from the amoeba to Einstein, organisms are in the business of formulating and testing hypotheses (Popper, 1972). Are the social science and the biological perspectives on knowledge incommensurable in a Kuhnian sense (Kuhn, 1962)? Or can they be reconciled? Our hypothesis is that through learning, biological entities transform capacities into objects and objects into capacities. We thus take learning—in its context of organization—to be the key to any reconciliation between them.

Stewart and Cohen’s concept of extelligence illustrates how objects generate capacities (Stewart & Cohen, 1997): by blurring their boundaries with their context. Extelligence, then, can be thought of as a relational property that links a system at a given level with its proximate environment at that same level, i.e., with its context. There is extelligence at the level of the individual agents within an organization—the organization itself then becomes a form of extelligence at this level. And there is also extelligence at the level of the organization itself—here, markets and other institutions now constitute extelligences for it.

So how, then, should we consider the term “organization”? Within each discipline, the meaning of the term has been evolving over the years.


The challenge of organizing was felt by political units long before it was felt by economic ones. The modern nation state would be unthinkable without a significant increase in organizational capacity. The rapid spread of education and literacy in western societies over the last three centuries allowed a phenomenal growth in the organs of modern governments and in their ability to monopolize the means of coercion and of administration over increasingly large tracts of territory. Within the modern nation state, coercion and administration have become inversely related: control by force gradually gives way to control by regulation, i.e., over time, information and intelligence substitute for energy, both constructive and destructive. But increases in the size of administrative units spawn impersonal bureaucracies—a shift from community to organization, from Gemeinschaft to Gesellschaft (Tönnies, 1955). The core values of these new entities suggest that bureaucracies are efficient machines in the service of the state. They have no goals of their own. They are means rather than ends (Weber, 1978; Elias, 1939).

With the advent of the railway and the telegraph in the second half of the nineteenth century—and later that of the telephone—the focus of economic organizations was still primarily on the transportation and distribution of physical goods. In the space of two decades, in a number of industries, the US evolved from a collection of disjointed regional markets into a single integrated national market. The growth of the modern corporation in the US at the end of the nineteenth century was a natural consequence of these technological changes. Yet, the sheer size of these giant firms posed the problem of managerial coordination in an acute way. Timetables had to be met, production scheduled, inventories monitored, etc. Unsurprisingly, the first modern managers were production engineers, and their prime concern was the planning and control of productive activity. Organization, then, was first and foremost the organization of production (Chandler, 1962), and the planning and control processes through which production was organized were akin to—to use a biological term—a nervous system.

The giant corporation grew through a process of differentiation and integration of increasingly specialized functions (Lawrence & Lorsch, 1967). Emile Durkheim’s distinction between organic and mechanical solidarity is relevant to this process (Durkheim, 1933). Durkheim was a turn-of-the-century French sociologist who claimed that in premodern societies, social solidarity was “mechanical” in nature, i.e., it bound together undifferentiated agents into a homogeneous whole. In modern industrial societies, by contrast, solidarity has become “organic,” i.e., because of the division of labor, agents are now specialized, so that where bonding occurs, it now takes place between functionally differentiated units. The first type of solidarity generates simple social objects, each with a “nervous system,” whereas the second generates complex social systems, and is much more like an ecosystem of diverse creatures;1 see below. This is reminiscent of Toffler’s First Wave and Second Wave; it is his Third Wave (1980) that matches our later informational concerns.

The implicit model that underpinned the concept of early twentiethcentury organizations—both state and nonstate—was drawn from a nineteenth-century physics deeply committed to a mechanical philosophy. With large-scale production, the concerns were focused on energy expenditures and achieving efficiency, i.e., “work put in” over “work got out.” The aim was to minimize the energy expenditures needed—animate or inanimate—to obtain a given result. “Second Law of Thermodynamics” thinking dominated: you couldn’t win, you were doing well to break even! In the “machine bureaucracies” thus created (Mintzberg & Waters, 1985), good management was first and foremost the management of costs, and cost accountants reigned supreme. They even had their own god: Frederick Taylor. Note that the reduction of input was seen, under the eye of the cost accountant, to be at least as virtuous an activity as increasing the output.

Associating the efficient use of machines with effective information flows and feedback relationships made industrial managers particularly receptive to cybernetic concepts when these first appeared in the 1940s and 1950s (Wiener, 1962; Ashby, 1954). Yet, instead of treating their organizations as open systems—as was increasingly being done in biology (see below)—industrial managers were continually trying to close them up in order to assert managerial control. In other words, they were constantly pushing the organization-as-machine model rather than the organizationas-living-thing model. They were trying, in a way, to emulate closed systems by reducing their inputs, improving “efficiency,” not efficacy. And even when the open-systems perspective finally took root—in the 1960s and 1970s—it promoted the view of organization-as-individual-organism rather than as anything larger or more complex. Organism here equaled firm (Burns & Stalker, 1960). Serebriakoff (1975) coined the word “org” specifically to refer to organism and to organization, with both having a “brain” between a sensorium and a motorium, with external feedbacks at least.

Open systems theory introduced a biological perspective to management. It naturally produced its own tensions. Organization theorists found themselves pulled in opposite directions by the Newtonian traditions that still pervaded economics and the Heraclitian traditions of biology. On the one hand, economists have never been particularly concerned with organizational matters. They have typically treated firms as atoms that collide competitively with each other in markets. What goes on inside firms—internal organization—is of no particular interest to them. Taking it as an axiom of their system that firms can be treated as if they behaved rationally, they can afford to ignore the complexities of what goes on within their organizational boundaries as epiphenomenal to their concerns. Rational firms will be driven to respond in identical fashion to a given set of external forces. (This had many prejudices in common with the Skinner behaviorist psychology, where organisms—and children— were considered as if there was no internal structure, simply the input/output system.)

On the other hand, and drawing on biological analogies, organization theorists had become increasingly uncomfortable with this simple mechanical representation of economic organizations. Looking inside firms, they sensed a degree of complexity and novelty in organizational processes that constantly challenged the toy-world models bequeathed to them by economists. Equally, the psychologists of the 1970s repudiated Skinner.

The massive densification of communications that has taken place over the past few decades has gradually eroded that facile distinction between “internal” and “external” organization that allowed economists to demarcate themselves from organizational theorists. External organization may have been treatable by the economists’ analytical tools. It could be managed through the impersonal and mechanical operations of the “market mechanism.” Internal organization, however, was too complex for that. It required managerial coordination (Coase, 1937; Williamson, 1975). Furthermore, with fast, low-cost modern communications technologies, external organization has now become as complex as internal organization. If 100 years ago the railway and the telegraph initially created islands of complexity in the form of the modern corporation, the internet is today extending this complexity and making it ubiquitous.

The rapid growth of interorganizational networking made possible by the arrival of the internet has promoted an ecological perspective on economic/business organization, as well as fostering a growing interest in complex systems. Organizations are now viewed less as tightly coupled objects than as loosely coupled systems in interaction. This shift in perspective clearly challenges the traditional assumption that we can associate economic organizations with bounded entities called firms. It has always been taken for granted that firms were both instruments of production and of distribution. But they turn out, on closer inspection, to be only instruments of distribution—a ranked structure of claims to the output of production.2 With the spread of outsourcing, downsizing, and strategic alliancing, productive organization today reaches out far beyond the boundaries of the firm as conventionally understood. The distinction between internal and external organization has now not so much disappeared as become intractably fuzzy.

Both social and biological organizations first have to construe their environments in order to act in them. Each does it in its own way. What it is to be a frog (Lettvin et al., 1959) thus differs significantly from what it is to be a bat (Nagel, 1989); frogs and bats construe their relevant environments differently. So do Microsoft and the Birmingham city authorities in the UK. We are here dealing with problems of representation and of meaning. It may be argued that complex living systems—and, more controversially, some complex nonliving systems—do not react directly to events but to their internal representations of events. The focus thus shifts imperceptibly to issues of knowledge and the emergence of patterns, i.e., the organization of information to generate plausible representations of an environment as the basis of action. Enacting such representations subsequently shapes an organization’s capacities and hence our conception of what it should or could be.3 When Wittgenstein (1968) refers to “forms of life,” is he not simply referring to the way that a capacity for having or generating knowledge is constrained by organization, biological or otherwise?

We should keep in mind, for organisms as well as organizations, that the context is not an unchanging “environment,” which the internal knowledge can assume remains the same. The structure of the effective context is determined recursively, like rivers carving their way across a landscape and changing its geography, then being guided by the new topography. Tadpoles see a different landscape from frogs, but themselves change the landscape. A puppy sees your home differently from the dog it becomes, but the dog’s life is modified by what it did as a puppy. Equally, the existence of a company in the “industrial ecology” should now be seen as a modifier of the very environment which that company sees as its context. This is new thinking whose characteristic word—and thought—is recursion.

Organizations, then, bring forth their worlds. Their knowledge is as much about the possible worlds that they are capable of construing as about the probable worlds to which they need to respond if they are to survive in the short term. They therefore act more in a matrix of plausibilities than of certainties. This is because, like biblical prophets, they are caught by their own forecasts and must look at the phase space of possibilities for the future; they must, now, have grown out of the “five-year plan” mentality of the 1930s. The complex adaptive systems approach is well aligned with this perspective. It offers us a much more provisional and potentially richer conception of what organization is about. It makes us less disposed to ontologize traditional forms of organization and more inclined to tolerate a “Cambrian” explosion of organizational possibilities (Gould, 1990).


Has biology experienced the same transition as did management from tightly coupled conceptions of organization to more loosely coupled ones? Has it moved from consideration of objects, to processes, to complex systems thinking? What have been the consequences? Has there also been a shift from an energy-based view of biological processes to an information-based one?

This change of perspective in thinking about human cultural organizations parallels earlier, arguably homologous, changes in biological thinking. Here are some examples. In the 1930s to 1950s, anatomy—a science of objects and their relationships—was treated as a major focus in schools and in medical education in general, with dissection as the central teaching exercise and a significant tool in research. This was replaced by physiology—a science of processes—in the 1950s to 1970s. Physiology started with closed-system, equilibrium assumptions (homeostasis), but then moved into the 1980s with progressively open-systems thinking regulated by internal feedbacks.

Endocrinology started in the 1920s with lists of hormones: until the 1970s, measures of “amounts” such as blood levels of oestrogens or highdensity lipids, for example, were the usual clinical parameters. After the 1980s, endocrine questions began to mesh in with more complex questions about feedback relationships. Today, endocrinologists have developed a vast list of chemical messengers, whose amounts have become less important than their timing and their responsivity within complex interactive systems.

Embryology in the 1920s was mostly descriptive; it then became progressively more interactive in outlook, looking to the organism developing in context, and regulating not its state (homeostasis) but its developmental path (homeorhesis). From this it was but a step to the epigenetic view, i.e., an organism-level trajectory operating in context.

Ecology was originally made up of object-oriented lists of species and numbers. During the course of the twentieth century, a more interactive and open-systems style of thinking gradually led to the nonequilibrium ecosystems perspective that we teach and manage today, drawing on progressively more complex-systems models whose mathematical basis includes bounded chaos. “Balance of nature” thinking finally died among professional ecologists in about 1985, leading to a dissonance between folk ecologists and conservationists that somewhat resembles that between “command-and-control” and “open-systems” managers in economic organizations.

In both the biological and the managerial disciplines, one sees two significant shifts: from objects to interaction between objects; and from objects as things to objects as spatio-temporal states in wider processes. In each of the two shifts, the time dimension looms larger. In both disciplines, one is now moving toward the study of interacting processes; in other words, toward the study of complex, possibly adaptive systems. The two shifts in our thinking—from objects to interactions and from objects to spatio-temporal states—vastly increase the degrees of freedom of what one understands by an organization, as well as the scope for creative and emergent processes to drive its evolution and development. This introduces the issue of organizational autonomy, which differs in substantial ways from Maturana and Varela’s concept of biological autonomy and autopoietic closure.4 It also moves us ever further away from the concept of the organization as a machine that is first “designed” from the outside and then externally directed, whether by William Paley’s watchmaker or by some other first-mover. To some extent, it also moves us away from the simplistic biology promoted by the popularizers of a DNA-driven world. Dawkins’ The Selfish Gene and The Blind Watchmaker return readers to a mechanical, object-centered view that is not widely shared by today’s working biologists, in the (mistaken) belief that it is a simpler view more congenial to the readers.

The ways of looking at the evolution of reproductive systems that were pioneered by Waddington in the 1960s (Waddington, 1957) are now becoming popular among biologists, for they fit this new-style thinking (but are not easy to meld with the DNA-centered, organisms-are-justDNA-writ-large approach so common in popular books and the media). These epigenetic landscape models and their kin aid understanding of evolving lineages by presenting each development as a series of balls rolling down an “epigenetic landscape” of hills and valleys; where the balls end up defines the resulting organism. The useful element of this model is the determination of what determines the topography of these successive landscapes. Think of each developing organism’s landscape as an elastic sheet. The gene system pulls from underneath, a series of tangled threads each attached separately to the sheet (for few genes have simple, single effects), while the environment pushes from above—think of a hand whose fingers are resting on the sheet. As each generation of organisms is produced, its existence modifies the shape of the landscape for subsequent generations, so different genetic patterns are selected.

This recursive genetic malleability is alien to the Mendelian, purely mutation-based explanations of evolution. Only a small proportion of organisms grow up to become breeders, and it is these which have recombined the parental genes in useful ways. Nearly all evolution proceeds by rearrangement of the very diverse genetics of natural populations. Only in laboratories, in folk biology, and in some science-fiction films are new mutations taken to be the sole source of variability.

At the scale of human organizations, our questions have also gradually shifted from “What is the secret process that a particular firm uses to remain successful?”—a “mutational” approach—to “How has this organization promoted a responsiveness to changes in the market that it has itself caused in order to remain successful?”—an evolutionary strategy used by Darwin’s finches.


Practice, as always, lags behind. Our perceptual apparatus is more comfortable resolving the world into objects than into processes. Objects have a lower dimensionality than do processes. “Folk” conceptions of organization thus reflect the fact that “being” is an altogether easier concept to deal with than “becoming” (Juarrero, 1999), i.e., we cannot easily cope with games of chess where the rules evolve or keep changing. For everyday purposes, stable objects tax our limited data-processing capacities less by sparing us the need to deal with complex dynamics. Nevertheless, given the need to cope with ever-increasing turbulence in our environment, our new concepts of organization are gradually moving away from “folk” approaches toward something more sophisticated. Being, as Parmenides pointed out, is ideally surprise free. It is therefore boring as well as misleading. Becoming, by contrast, is Heraclitian; as flux, its trajectory is open ended. More risky than being, to be sure, but in a turbulent world, also both more realistic and more fun.


In discussing how biologists and organization scientists use the term “organization” in their reasoning, we first need briefly to distinguish between reasoning by means of metaphor, by means of analogy, and by abstraction. How do these three forms of reasoning differ? Simplifying somewhat, we might say the following:

  • Reasoning by metaphor treats things that are different as if they were similar in one significant respect that is left largely implicit. Metaphors achieve their effects by connoting rather than denoting.

  • Reasoning by analogy treats things that are different as if they were the same in a number of significant respects that are more rigorously defined than in the case of the metaphor. Analogies denote rather than connote.

  • Reasoning by abstraction treats things that are different as if they were the same in all significant respects (Dretske, 1981). As we move from metaphor to abstraction, our reasoning shifts from the poetic to the propositional.

There are seductive metaphors that link human activities and the biological world, built into our everyday language and therefore difficult to analyze. We talk, for example, of the head of the firm, the body of the church, the long arm of the law, the minister’s right-hand man, and so on. In the middle ages the “body corporate” made its appearance. It allowed a monarch, for example, to treat a group of people sharing a common interest or concern as if it were a single unified entity. Charters of incorporation, given to craft guilds or city corporations, allowed them to relate to the crown and other parties as a single body (Turnbull, 1997). The goals and interests shared by members of the corporation justified the assumption that this body had a mind of its own and and that it could act rationally with respect to such goals and interests. It was thus endowed with a “legal personality;” the joint-stock companies of the nineteenth century were built on the same assumption.

The dangers of metaphorical reasoning are well known. They provide tools for explanation, giving us insights rather than understanding, and insights can often prove illusory. They must be considered points of departure for the reasoning process rather than points of arrival.

As epistemological strategies, analogies fare somewhat better; but how useful are they? Most such analogies are useful in illustrating a particular point or illuminating a specific problem. Let us get back to biological analogies of social entities for our examples, particularly misleading ones. Usually the comparison is with the “folk” idea, and the biology is very different—we don’t accept the argument that if the “folk” idea is common to explainer and explainee, it doesn’t matter what the reality is. Comparison with a cell, for example, is useless if you get the cell all wrong because you’re making the comparison with the “lies-tochildren version” (Stewart & Cohen, 1997) from the elementary biology textbook. On the other hand, some published analogies have been useful and fairly true to the realities of biology. Victor Serebriakoff’s “orgs” (1975), for example, were a good way of comparing sensory and motor physiologies of organisms and factories, and Popper’s view of knowledge as a capacity to act (“from the amoeba to Einstein”) is also a useful image for both economists and biologists.

The use of analogies presents two dangers. Either superficial characters are being analogized, as in the biological arena—for example, we might compare zebras and giraffes and tigers, and conclude “camouflage features.” The problem then arises when we move out along the yellowand-black axis and find wasps, whose pattern serves a totally different function, being a conspicuous warning. Alternatively, perhaps, we might explore the deeper analogy between carnivores that have their eyes more or less at the front where they can rangefind and concentrate on their prey, with the open-field herbivores that are their prey and that have eyes on the sides, where they can observe almost 360 degrees. Nice idea, but again there are some contradictions: one must ask, for example, if herbivores have any special way to look close and down at what they are eating, and then again look at wasps, which are carnivores with 360-degree eyes.

Every analogy is biased by the theories of the two items, or the two processes, that are being compared. Yet, we do feel that we can do more, after we have worked out that a thermostat is like a ballcock valve on a cistern is like an essay returned with tutor’s comments. We can say “negative feedback.” Our query in this article relates to the feeling that we can do more when we say that, for example, cybernetics underlies both industries and ecologies. With the growth of the managerial sciences and particularly the development of open-systems theory, it began to look as if one could go beyond metaphorical references to biology and reason more rigorously by analogy. Are we today able to move beyond analogy and further toward an abstract concept of organization—one that would be shared by the biological and the social sciences?

The cybernetic insight was an important one, but it remains limited in scope. We have moved from object to process, to feedback among processes (cybernetics and homeostasis), and then to phase spaces and epigenetic landscapes (homeorhesis and phase spaces). Indeed, we have moved on further, to the recursional processes that can be discerned in phase space (evolutionary fitness landscapes, cubes of Information Space). Now we can imagine that one company (or ecosystem) can take a particular developmental trajectory, adapting to a changing context as it progresses, even as it shapes it.

Knowledge involves a selective representation of important “features” (Stewart & Cohen, 1997) of an external or internal environment for the purposes of choosing and acting. But choosing and acting do not of themselves imply a living entity. Any feedback mechanism requires some degree of representation; indeed, that is precisely what the word “representation” means—feedback! The structure of a thermostat, for example, includes a thermometer (bimetal strip) that generates representations for itself in that sense. Living things, however, usually have more complex representations and can do more things with them than can inanimate things like thermostats. Beyond a certain level of complexity, the choosing and acting associated with representations can lead to agency.

One more word about the limitations of cybernetics. Like many of the icons that became a source of analogies for managers—the second law of thermodynamics, homeostasis, and the balance of nature—cybernetics was the engineering manifestation of a reductionist strategy. It was an attempt to reduce the living to the mechanical. It was only ever interested in the feedback processes of simple systems, or those of systems that it could eventually simplify. It neither sought nor could cope with complex and interwoven feedback mechanisms, those that could give rise to emergent properties. Cybernetics was not interested in emergence; it was interested in control (Wiener, 1962). Its goal was the production of servomechanisms, i.e., of agency at the most primitive level. So, if organizational thinking traces a shift from objects to processes—i.e., interaction between objects—cybernetics looks at the subset of interactions that is characterized by feedback. But, as the feedback loops increase in density, the interactions generate a level of complexity that puts them beyond the reach of cybernetics, rooted as it is in the concept of mechanism (Wiener, 1962).

Emergence then comes to the fore as a phenomenon in its own right. Being a generator of hierarchies, emergence challenges the reductionist strategies on which the concept of mechanism was originally built. Simply put, reductionism creates simple objects, whereas emergence creates complex objects. Social and biological organizations are instances of complex objects, the outcome of dense interwoven processes unfolding over time.


With the time dimension now coming into the picture, it becomes necessary to introduce the idea of irreversible processes. This is what the second law of thermodynamics is all about. But the second law turns out to be limited. Inheriting as it does the physicist’s closed-system perspective, it implicitly equates irreversible processes with degenerative processes. Yet, irreversibility in time refers to nothing more than the system’s loss of memory, i.e., its access to past states. It kicks over its tracks and hence cannot retrace its steps. The assumption that this automatically leads to degeneration or disorder is unwarranted. Through the phenomenon of emergence, it can also lead to order (Prigogine & Stengers, 1984; Nicolis & Prigogine, 1989). In this sense, emergence is the antithesis of entropy. Emergence does not actually violate the second law —it respects the idea of irreversibility—but rather gives rise to the idea that temporal processes can be the source of a local order. This order may well be paid for in the coin of entropy generated somewhere outside the system, something that still presents difficulties for many physicists who then have to deal with the phenomenon of negative entropy (Schrödinger, 1967). Such a local order must be considered ontologically privileged, i.e., it skews outcomes in favor of certain states at the expense of competing alternatives. Stewart and Cohen discuss the concept of privilege5 in the biological sphere in their book Figments of Reality. We here extend the concept of privilege even further, to all physical processes.

In sum, we argue that emergence and entropy are two sides of the same thermodynamic coin. Both are characterized by irreversibility, but only in the case of entropy does irreversibility lead to increasing disorder. In the case of emergence, it leads to increasing and ordered complexity. Both social and biological organizations are subject to these thermodynamic effects. They are privileged sites of what Schumpeter (1961) labeled “creative destruction.” At such privileged sites, we witness a replay of the ancient battle between Parminides and Heraclitus, that is, between the conflicting forces of stability and instability. According to Alicia Juarrero, Parminides won (Juarrero, 1999). Yet, with the spread of the internet and the evolution of ever more turbulent organizational environments, it may soon be time for Heraclitus to make a comeback.

Emergence is the ultimate Heraclitean process. It is a generator of ontological hierarchies. It first builds these from the bottom up and then these, in turn, control the resulting articulated systems from the top down. There is a range of levels over which comparisons are often made between biological hierarchies and social ones. Thus, evolution, ecologies, colonies, bodies, cells, bacteria, viruses, prions, etc. find echoes in our thinking on human organizations at the level of nations (ecologies), industries (colonies), firms (bodies), individual divisions or departments (cells), bacteria (employees, ugh!), communications (viruses?), prions (memes?).

At each level in a system’s evolving organizational hierarchy there is the challenge of establishing a unit of agency and its structure, as well as its characteristic processes and the constraints that act on these. The processes and constraints then change the rules for the system’s next iteration. Recursion produces changes which are progressive, which build difference each time around. Emergence is one possible outcome— entropy being the other—when the system’s history is forgotten: whatever information might have been available as to its causal structure is lost to view. One must bear in mind, however, that the view in question is only ever the observer’s. Emergence is an observer-dependent phenomenon, discernible only from outside the system. Yet, remove the observer and you get what Thomas Nagel has labeled “the view from nowhere” (Nagel, 1989)—we must observe to discern emergence, because many systems, like a swimmer in a current, cannot “notice” their context changing.

Emergence exploits the contextual properties of structures and/or systems. It also exploits the degrees of freedom available to systems. Where the elements of a system are tightly coupled, for example, the scope for variation—and hence for emergent processes to take root—is quite limited (Boisot, 1998). It resides, if anywhere, in the combinatorial potential of its constituent elements: both the properties and identity of the system are largely determined by the combinations that it can achieve. Where, on the other hand, the system is characterized by a loose coupling of its elements, we get not only combinatorial power, but also behaviors, that is, combinations that vary over time. Thus the binding strength of subatomic particles is so high that not only does it allow little combinatorial potential and hence variation—they are too tightly coupled for that—but it also allows little in the way of behaviors. The weaker binding strength of atoms, by contrast, gives us all the potential combinations available to us through chemistry, together with a variety of behaviors that can now take place within and between molecules. Some of these behaviors are complex enough to allow organic molecules and autocatalytic processes to emerge (Kauffman, 1993, 1994). Finally, when transmitted information, rather than energy, becomes the primary binding agent—as it is between autonomous organisms that have representations of each other and can signal to each other—potential variations register increasingly, not at the level of what the thing is, but of what the thing does, i.e., as choices predicated on sequences of behaviors. Iterative behavior evolves, too.

We can frame the tight/loose coupling issue as a relationship between two kinds of resource that are needed to establish and then stabilize interactions between objects: binding energy and what we will call binding information. Hydrogen and oxygen rely primarily on the first kind of energy, our immune system relies primarily on the second. Energy and information are both required to keep a system together, but in different mixes. Thus, for example, whereas simple systems will rely mostly on binding energy to keep themselves together, complex systems will look to binding information to maintain their integrity. This can be illustrated by means of a simple graph. We can place the system’s binding energy on the x axis and its binding information on the y axis. In the scheme just outlined, the amounts of binding energy and binding information required to keep the system together are inversely related. With very high binding energies, the elements of the system covary, so that knowledge of the state of one element is sufficient to give you knowledge of the state of the system as a whole. Little binding information is needed. As the binding energy decreases, however, the number of possible configurations that the system can adopt goes up and so does its potential information content. It then requires a larger amount of binding (constraining) information to keep it within a range of viable configurations. The graph allows us to conceptualize the distinction between tightly coupled and loosely coupled systems in information terms. It also allows us to distinguish systems from aggregations or piles. What we call objects, in effect, correspond to tightly coupled systems, whereas what in this article we refer to as organizations correspond to loosely coupled systems.6

Many problems in the history of organizations stem from attempts to treat organizations as if they were objects. We submit that the reason for this has to do with the cognitive challenge of dealing with loosely coupled systems. Owing to their higher degrees of freedom, they are inherently more complex than objects, and such complexity is not intuitively accessible. Organizations tax our powers of abstraction to a much higher degree than objects.

We can perhaps now better understand the nature of emergence as ontological privilege. As the number of possible states that a system can adopt goes up, it becomes increasingly difficult to understand why it first “chooses” certain states over others—i.e., it “bifurcates”—and then commits to these by subsequently self-organizing around them. We cover up our incomprehension by saying that novelty has been introduced into the system. In effect, though, the system’s “choices” reflect nothing more than an extreme sensitivity to the numberless discontinuities to which its own complexity give rise.


How far do the above remarks apply to both biology and the social sciences? To what extent do they suggest a set of abstract organizational principles common to both? We do not attempt to give a definite reply to these questions. Instead, in the form of bullet points, we offer a number of pointers:

  • A set of interrelated elements exhibit both recurrence and stability in their relationships; complexity here is but a measure of the organization or disorganization present in such relationships. Binding energy and binding information both contribute to recurrence and stability and, in so doing, they foster emergent processes.

  • In both biological and social systems, hierarchies are bottom-up emergent outcomes of complex organization. In both cases, however, once they have been created, hierarchies act in a top-down fashion to constrain and thus to further organize the system. This is also true of social networks. Sooner or later, they exhibit hierarchy, and such hierarchy then helps to shape the organizational behavior of actors in the network. Iterative behavior will change the rules through time.

  • In such a description, the antithetical relationship between entropy and emergence needs to be clarified. Both reflect the irreversible nature of the passage of time and the erasure of memory within a system. But, as Prigogine has shown in his work on dissipative structures, entropy and emergence lead off in opposite directions: the first toward increasing disorder, the second toward increasing order. Entropy has largely been absent from the discourses on social and economic organization, and emergence has only recently been admitted to it, but both need to be treated together as two sides of the same coin. The idea of ontological privilege depends on it.

  • In biology, the entropy concept has been applied both to energy processes (Brooks & Wiley, 1986) and to information processes (Atlan, 1979; Ayres, 1994; Küppers, 1990). We think that these approaches have been unnecessarily tethered to second-law degenerative ideas. It follows from what we have said above that the concept of constructive emergence should also be applied to energy and information processes respectively.

  • Above a certain level of complexity, biological systems have ways of representing their environment to themselves. They now respond to their environment partly directly and partly indirectly via the representations that they have of it. This will be equally true of social organizations. In the case of social organizations, however, there is a choice of representations to which one can respond. That is to say, representations compete with each other in a meme-like fashion for ascendency (Dawkins, 1986; Blackmore, 1999). Which representation is finally selected reflects the outcome of a political process. There is a seductive analogy here with Dennett’s “pandemonium” (Dennett, 1997).

  • In both cases, these representations form one of the terms of what we might call a “comparator;” the other term is given by actual outcomes as they occur in the real world. Serebriakoff (1975) developed the ideas of sensorium and motorium for all “orgs,” and the brain was a “comparator” so that behavior could become appropriate. The comparator shapes expectations and dispositions to react and behave in particular ways. It is the mismatch between expectations and experienced outcomes, whether recorded biologically or cognitively, that shapes behavior. In economic organizations, planning and control systems should fulfill the function of comparators.

  • A comparator is thus an embodiment of the knowledge held by the organization. It is also the main source of agency in the system. But agency implies choice and choice in turn implies alternative possible states of the system. Thus, agency only makes sense in complex adaptive systems that are capable of generating multiple alternative representations for themselves. Only such systems are capable of agency.

  • So we see that knowledge and agency are closely related concepts. In some deep sense, agency is predicated on the possession of knowledge, taken here to mean a disposition to act in one particular way when faced with alternative possibilites. We do not need to get “epistemological” about this. We could simply take knowledge to be that subset of your beliefs on which you are prepared to act.

  • We can then hypothesize that, to the extent that structure conditions a disposition to act, structure is itself an embodiment of knowledge. The implication is that any structured organization is also an embodiment of knowledge. Not only does an organization have knowledge; an organization is knowledge. The analogy with organisms is again helpful here: an organism is a process, a capacity to act that imposes its order locally—a whirlpool or a fountain rather than a rock.

  • If we follow Brooks and Wiley in arguing that biological systems are concerned to minimize their rate of entropy production per unit of work performed (Brooks & Wiley, 1986), we can see something similar happening in social and economic organizations as they relentlessly seek out greater efficiencies. This is good 1960s thinking. It was applied both to their energy processes and to their information processes. Yet, we now know that a quest to minimize entropy production must be balanced out with a need to handle variety, as iteration proceeds and the organization changes its own context. Emergence builds on such variety; a system that was totally efficiency oriented could never evolve. Thus, ultimately, it is the turbulence with which a system has to deal that determines how it will balance out the competing claims of entropic and emergent processes. (The cost accountant is precisely the spanner in the recursive works!)

  • Minimizing entropy production in both biological and social systems is an economizing activity. It takes place in two steps: first, substitute data processing for energy processing through cumulative learning; second, reduce the data-processing load through acts of data structuring. Boisot (1998) takes data structuring to consist of codifications and abstractions. This second step requires insight, which is itself the outcome of an emergent process. Once more, we find entropy and emergence to be inextricably intertwined.


Until the last two decades of the twentieth century, economists remained tethered to an energy metaphor drawn from nineteenth-century physics (Mirowski, 1989). Alfred Marshall had hinted that biology would be a more appropriate source of concepts for economists than was physics, but the hint was never taken up. Today, evolutionary economics has become a branch of economics in its own right (Vromen, 1995). The study of human organizations has aligned with biological thinking earlier and more extensively than has physics (Aldrich, 1999). But has it moved beyond metaphor?

Physicists, while trying to create a toy universe based on linear mathematical thinking, committed themselves to a Boltzmann heat-engine thermodynamics that led to a heat-death picture of the universe and an interpretation of living things as feeding on negentropy. Yet, this was but one choice open to them. If, for example, they had included gravity as well as perfectly elastic billiard balls in their thinking, their closed system would have gone up in order instead of down! Early conceptions of organisms as machines (Descartes, 1989) reflected the billiard ball view, but most biologists have now moved beyond it.

Von Bertalanffy developed systems theory mainly to deal with homeostasis in physiology. He realised that feedback processes implied that time had a direction to it, and that physical “laws,” contra Newtonian systems, could not read back and forth (compare his position with that of Schumpeter, e.g., in Boisot, 1998). From Newton you could explain orbits and predict Uranus from anomalies in Neptune. From endocrinology you could explain certain behaviors, especially reproductive behaviors and pathologies, and occasionally predict: the hormone activin, for example, was successfully predicted, but it excited less media attention than the discovery of Uranus. From Darwin, however, although you could explain Galapagos finch species, you could not predict the appearance or the characteristics of another one. Understanding and explaining carried both different meanings and different requirements for physicists on the one hand, and for biologists, geologists, or astrophysicists on the other. The former are concerned to treat events that are ostensibly different as if they were the same, i.e., they abstract (Dretske, 1981; Boisot, 1998). The latter are more comfortable taking events that ostensibly look the same—in that sense, they also rely on abstractions—but then focusing on what makes them unique.

Biologists, in sum, have come to realize that individuals, including humans, all differ genetically by quite a lot: not only does a species not have a single genetic blueprint, but all its individuals work in different ways. They are not Ford Model Ts but handmade cars, each adjusted in a different way to give collectively much the same results in spite of the presence of good or bad mutations. Likewise, at the level of human organizations, it became apparent that standardized blueprints for company startups did not work as well as strategies that led to differences. The resulting heterogeneity effectively reduced competition and improved a startup’s chances of survival.

The new sciences of complexity explain but do not predict. Prediction, as conventionally understood, is beyond their reach. Such a limitation is inherent in their subject matter. Yet, can a science that does not predict be useful? The “failure” of complexity sciences to predict is due at least in part to their dependence on a loose coupling among the elements of the systems that they study for versatility in behavior. Couplings must be loose enough to allow different—and sometimes novel—behaviors to appear at successive iterations. Hence, no prediction. Such loosely coupled systems lead to evolving and emergent processes because their combinatorial potential is greater.

Complexity sciences have self-organizing processes as one of their central concerns. With the passage of time, the universe tends to complicate itself through a process of symmetry breaking, changing its own rules as it does so (Kauffman, 2000). This “evolution” of purely physical processes has led physicists to start thinking more like biologists. Yet, what applies to physical processes also seems to apply with even more force to social ones. As organizations evolve, they become more complex, i.e., more informative. Organizations thus tend to metamorphose over time in ways that cannot be predicted ex ante. Indeed, some would argue that we are witnessing precisely such a metamorphosis in the realm of social organizations as a result of the enhanced interconnectivity made possible by the internet.

The fact that both the “hardest” of the natural sciences and the “soft” social sciences are now both focusing on emergent processes and selforganization suggests that we are indeed gradually moving away from metaphor and toward a single science of organization. Biological forms of thought are thus increasing the size of their basin of attraction and metamorphosing as they do so.


  1. Mechanical and organic solidarity correspond to the distinction that ecologists draw between commensalism—i.e., competition and cooperation between similar organisms—and symbiosis—i.e., cooperation between dissimilar organisms. For a fuller discussion of the distinction, see Aldrich (1999).

  2. Marx, in the distinction that he drew between the “sphere of production” and the “sphere of exchange,” was making essentially the same point. He drew different conclusions, however, to those put forward here; see his Capital.

  3. Note here that the employees have a conception, an idea of what the organization should be; this may differ from the managers’ conception. And both probably differ from the consultant’s conception of what the organization should be, who probably sees a different context for it and so understands it differently.

  4. Maturana and Varela once more reproduce the separation between internal and external organization that economists once imposed on the social sciences. They thus promote the idea of an organization as an object with an inside and an outside. Organizational autonomy is maintained in part by closed boundaries. This will be true of some but not all organizational forms. In some cases, the integrity of an organizational form is achieved by an attractor exercising its influence over a field. Here, there is no clear boundary that separates the inside from the outside, only a gradient of field strength.

  5. The extended usage of this concept is introduced in Cohen (1977).

  6. Clearly, all “objects” are characterized by some measure of organization. We recognize that organization is a matter of degree. Some objects can be highly organized without constituting organizations for our purposes.