E:CO is republishing this paper in our Classical Paper section not so much because of its venerable age and influence—it hails from 1992, 24 years ago—but because of its incisiveness, insight, and acumen in examining a crucial issue whose importance has only increased with time, namely, complex systems as integrated wholes. Its author, the systems theorist Rod Swenson, comes from the field of ecological psychology, one of the many strands of systems-oriented disciplines which have come together in the formation and development of complexity science. Ecological psychology has not only offered an impressive and influential body of substantive research and theorizing which has undermined not a few traditional, misleading presumptions plaguing the fields of perception, cognition, and related areas but has played a key role in facilitating the changeover from systems theories per se to the complex systems perspective taken-up in complexity science.
Swenson’s paper has a two-fold aim. One is a trenchant critique of the theory of autopoiesis developed by the Chilean scientists Humberto Maturana and Francisco Varela in the late 1970s and thereafter enlivening research programs mainly in South America and later in the US. First devised within the context of theoretical biology as a way to buttress organic wholeness against its encroachment by simplistic information theoretic input/output models, autopoiesis was later applied to social structures and other arenas where the notions of closure and autonomy were relevant.
Swenson’s censure of autopoiesis is based on several factors. First there is a repudiation of what he sees as its deficient epistemological claims (particularly as propagated through the lens of Maturana’s surprisingly sophomoric epistemic framework). Second, there is the overemphasis, on the part of proponents of autopoiesis, on systemic "closure", Varela’s term for the opposite of the openness of information theoretic open systems models. Third, and at the center of Swenson’s own alternative formulation, was Maturana and Varela’s neglect of thermodynamics and dynamical systems’ explanations of systemic wholes observed. The purpose of this introduction is to provide additional historical and conceptual background to supplement Swenson’s own understanding of autopoiesis.
The origins of autopoiesis
The origin of autopoiesis stands at the confluence of several varied yet related streams of systems-based theorizing after WWII. One of the most important was the treatment of the idea of homeostasis, a central focus in cybernetics1. The most thorough and astute explication of homeostasis can be found in the work of the brilliant British psychiatrist, neuoroscientist, and mathematician W. Ross Ashby1,2,3 (see also the highly informative 4).
Not one to take popular notions for granted, Ashby proposed a mathematical formalism for what he called “The Homeostat”, a hypothetical machine (Ashby built an actual prototype based on a set theoretic axiomatic foundation) that he held provided a sufficient description of how a living organism could learn and adapt to a changing environment. His homeostat had a small number of essential variables serving to maintain its operation over a wide range of environmental conditions so that if the latter changed and thereby demanded a shift of these variables, a new higher level of the machine would be activated to reset lower level internal connections that would then reestablish equilibrium (that is, the invariant state operating within a range of pre-set variables around which the homeostat functioned). A typical example is a thermostat (in older times also called a servo-mechanism) which controls a home’s ventilation/heating/cooling system. The thermostat is initially set at some invariant temperature and then when the ambient temperature changes, mechanisms are activated to either cool or heat the room in order to regain the original state. In Ashby’s model of a homeostat, like the role of random mutations during evolution, if environmental changes prod the servo-mechanism to become activated resulting in a more functional condition then it can be said that the homeostat-controled system survives, otherwise it expires.
At its core, the homeostat seeks equilibrium which means it seeks to maintain its current equilibrium or unchanging functioning to the extent that if somehow knocked off its current mode, it will institute a series of operations to restore that equilibrium. Similar ideas on equilibrium-seeking systems were to be found in physics, engineering, physiology, economics (perhaps the discipline where it was formalized in the most arduous fashion) and social system dynamics.
In the social sciences, for instance, the social-psychologist Kurt Lewin, who had emigrated to the US to escape Nazi persecution, understood social systems in terms of a field of forces in equilibrium, so that any disturbance of the equilibrium force-field would be compensated by countervailing force5 . Similarly, in the field of sociology, Talcott Parsons and others understood social systems as large self-regulating, equilibrium-seeking systems6; an idea that went back to early notions idea of self-organization and self-regulation proposed by Kant, Schelling, and other founders of naturphilosophie in the eighteenth nineteenth centuries (see 7 on the fascinating history of self-organization).
Regarding Kant, whereas Goethe was to become known for his aesthetic/perceptual appreciation of organic wholes, Kant (cited in 8) had hoped to elucidate the unique nature of life as opposed to the inorganic by way of a self-referential causality, “It is necessary that its parts should combine to form the unity of a whole by being reciprocally causes and effects of each others form. For only in this way is it possible for the idea of the whole ... [to] determine the form and combination of the parts (my emphasis)”. Juarerro-Roque (citing Kant’s Critique of Judgment) has pointed out that for Kant an “intrinsic” physical end was “both cause and effect of itself,” for example, a tree which, “...in the genus, now as effect, now as cause, continually generated from itself and likewise generating itself, it preserves itself generically.” It was the nature of this circularly causal structure which supported Kant in claiming that biology could not be reduced to physics since the organization of the organic whole, where end and means coincide, must be accepted as the "primary given starting point of investigation within the organic realm."
Kant turned to a tree to illustrate what he meant by the circular causality of living organisms (see 9; Moss is another biologist turned philosopher of biology). A tree showed circular causality in three ways: first, in terms of speciation, the tree is both cause and effect of other trees of its kind; second, the tree, in causing an assimilation of nutrients to produce its required constituents, at the same time causes the chemical changes that these nutrients undergo in becoming the “matter” of the tree; third, the parts of the tree are reciprocally dependent on each other, e.g., shoots dependent on roots, shoots on roots.
The living organism was self-referentially and thereby circularly causal in yet another way. According to Rae, the “inertia” of matter referred a lack of life, “life” being the capacity of a substance to determine itself by acting from an internal principle10. However, Rae points out that, "To say that every change in matter must have an external cause is, for Kant, precisely to deny that it is self-determining." In other words, it is the self-referential, circular causality inherent in life, particularly in the sense of organic wholes, which renders life as life. We’ll see that this circular formulation of what makes life life is not far from what the theory of autopoiesis was going to claim some two centuries later.
The self-referential, self-determining interpretation of organic wholeness was incorporated into theoretical biology during the nineteenth century, undergoing many modifications along the way, some of which did not stress the same themes found in naturphilosophie. The conceptual earthquake caused by the theory of evolution hit theoretic biology broadsides, yet the issue of organic wholeness in its self-referential form was not especially effected by it until the middle of the twentieth century when some theorists wound it up with cybernetics and information theoretic models. It is in opposition to the latter which is the context in which autopoiesis arose.
Cybernetics carries an equilibrium seeking perspective on to whatever arena it touches. In this respect, we find an emphasis on negative feedback as the countervailing operations which restore a system to equilibrium, i.e., a convergence back to equilibrium, whereas the opposite tendency to augment disturbances in the direction of divergence is termed positive feedback. It is really only a short step from conceiving a systems’ ability to restore equilibrium after being disturbed through the means of negative feedback loops to the full blown understanding of organic wholes as systems essentially constituted by self-and cross-referential causality.
But while this conceptual trend was being hashed out, another dilemma emerged by the very way convergence and divergence from equilibrium was being conceived. This was the dilemma of change of organic wholes, or how novelty in functionalities, capacities, and so forth could possibly come about. The horn of the dilemma had to do with how a system, basically functioning to restore equilibrium upon disturbance could ever allow something novel in? Wouldn’t the same mechanisms established to maintain the circularly closed integrity of the whole always seek to squash anything even remotely appearing to move the system off its invariant setting? And thus wouldn’t every tiny movement toward change be stamped out before it had any chance to establish new roots? It was questions like these, albeit maybe not expressed in such an explicit fashion, that penetrated theoretical biology as it became more and more a systems type of field of study and as circular causality rose to dominance as an explanatory construct.
It was the blossoming of cybernetics and artificial intelligence post WWII that brought the ideas of circular causality and self-reference into prominence. Of particular importance was the Macy Foundation Conferences on Cybernetics which gathered together many of the leading scientific lights of the day as a forum for the continuation and elaboration of their wartime efforts on intelligent machines, control systems, and the like1,11. The title of their first meeting in 1946 affords insight into the general direction taken by these cyberneticians: “Feedback Mechanisms and Circular Causal Systems in Biological and Social Systems,” a theme which remained in effect for the duration of the conferences until the last one in 1953. As pointed out by the polymath Gregory Bateson, who, as a contributor to many of these Macy Conferences helped introduce systems ideas into the social sciences, in the specific case of feedback, causal influence could be traced around a feedback circuit back through whatever position had been arbitrarily chosen as the starting point12. This circularity implied that any event in any position in the causal circuit of feedback would affect all other events at different positions, thus implying a self-referential structure since the causal influence was bent back on itself. Maturana himself was an important researchers among the so-called second order cyberneticians along with Heinz von Foerster and Gordon Pask and others. He is best known as the lead author of the paper, "What the frog's eye tells the frog's brain"13.
The steady state of General Systems Theory
The circularly causal, self-referential role of feedback cycles in biological systems was central in the cybernetics-influenced “general systems theory” (GST) developed by the German (and then American) embryologist Ludwig van Bertalanffy (another one of those profound humanists who were Nazis when it helped their career but later tried to squirm their way out of accusations of Nazism). It was the role of feedback loops as the conceptual foundation for organic wholeness that enabled GST to promote its anti-reductionist agenda without having to resort to any sort of vitalist supra-natural explanation. Jean Piaget, the celebrated Swiss developmental psychologist, much admired this aspect of Bertalanffy’s work believing it provided an authentic way station between mechanism and vitalism and thereby achieved what Emergent Evolutionism had only been able to promise, but ultimately couldn’t deliver because of what Piaget claimed was the latter movement’s overly “phenomenological and irrational” perspective14. It is hard to make much sense out of Piaget's criticism since "phenomenological" might be thought instead to be necessary feature of research and theorizing.
In GST, webs of interlocked feedback loops produce a steady state characterized by a capacity for the whole to restore itself after being disturbed by biochemical or tissue changes. This steady state was conceived as a dynamic or moving equilibrium in contrast to the static equilibrium of mechanical systems. As the goal towards which an organic system was directed by means of its feedback loops, the steady state was equivalent to the maintenance of the system’s integrity over time, a goal Bertalanffy described as equifinality meaning it was independent of initial conditions and could accordingly be reached from innumerable antecedent conditions a bit like the idea of an attractor. At the famous Alpbach conference in 1968 devoted to anti-reductionism, Bertalanffy boldy put reductionist and mechanistic science “on trial” for neglecting what he called the essential binding qualities he believed were necessary for understanding the functioning of parts within the integrated complexity of wholes15. Reductionist/mechanist explanations were secondarily imposed on a more primary network of interacting feedback loops in such a manner that there would not be any need for the former if suitable constraints were in place or emerged out of the latter. At the same time, the coherence supplied by the binding quality of feedback and its circular, self-referential causality also demarcated organic wholes against their environments.
Such wholes, seemingly closed with respect to their environments, were in actuality “open systems” since there is a continuous exchange with the environment: there is an input, a process of transformation and an output, all taking place within the context of anabolic and catabolic processes. They key was on understanding the equilibrium sought by such systems as a dynamic equilibrium, a steady state relying on a continual replenishment coming from exchanges between systems and their environments.
The issue of “open” vs “closed” systems at that time and continuing through Swenson’s paper even until contemporary theorizing became imbued with ambiguity and general confusion. Exactly how “open” did a system need to be if it was to simultaneously thrive on environmental exchanges while at the same time, as Piaget pointed out, make-up an inviolable arena between input and output? An organic whole required enough inviolability for it be a primary activity, not merely a reactive phase. Varela's closure, as offering greater autonomy and self-determination, became the keystone of autopoiesis unfortunately resulting in the kind of dreadful epistemology that Swenson strongly rejects. Moreover Swenson also had to grapple with this issue of closure in explaining the containment his approach required as well.
Weiss’ self-supporting arch
Paul Weiss was a well-known embryologist who was invited to the Macy conferences mentioned above (not to be confused with his contemporary name sake, the American philosopher Paul Weiss, whose work intersected in quite a few places–I also had conflated these two into one person, thinking it highly unlikely that two people of that same name co-existed with similar research programs at the same time).
The Paul Weiss we are talking about here is the one who came up with a different analogy for the self\referential structure of organic wholes: a self-supporting arch where each component could only be supported to the degree all components supported one another at the same time, i.e., "...a self-supporting structure can only exist in its entirety or not at all"(Weiss quoted in 16:166). The figure of a self-supporting arch, though, brings with it the enigma of how a self-referential whole could arise in the first place in an incremental fashion, the way evolution was supposed to operate, since it could only exist if all the parts were at once simultaneously supporting each other. This is a dilemma which plagues all attempts to understand the onset/origin of self-referential structures. Weiss himself didn’t believe there was a cogent story of how from the mechanism of parts alone there could emerge in an “integrated overall systemic order” of the whole. He compared it to the enigma of how a railroad on tracks could keep making progress towards its destination if the tracks suddenly came to an end in a wasteland. Weiss speculated that the system as an integrated whole gave “guiding cues” in the form of a “dynamic field structure of the total complex.” In fact, the concept of “fields” had been in vogue as a way for understanding referential wholes since at least the days of the electro-magnetic fields proposed by Faraday and mathematically formulated by Maxwell as well as the perceptual fields of Gestalt psychology then went all the way back to Goethe17.
Weiss was not as enamored as Bertalanffy with the idea of open systems, persistently criticizing such an approach as seeing systems merely as way stations for input to be transposed into output1. Again, one of the explanatory linchpins of the emphasis on closure which Varela stressed as a main characteristic of autopoietic systems was a strong repudiation of information theoretic input/output models (Henceforward, we can refer to autopoietic wholes in terms of referential closure since they exist at the intersection of cross and self-reference). Weiss, and later the theory of autopoiesis, suggested instead that input should be more accurately seen as operating somewhat like a “trigger” which selects from a variety of choice options for autonomous functioning. Foreshadowing the self-referential basis of autopoiesis, Weiss conceptualized referential wholes as requiring “the specific cooperation of their own terminal products.” Rejecting the holistic source of such terms for integrated wholes as “orgs” or “holons” which seemed to envision wholes as disembodied agencies, Weiss proposed instead a more action-oriented language to supplant the discourse of molecular biology and its program of genetic determinism.
Weiss offered a mathematical formulation for a whole (in particular a cell) in terms of how it systemically constrained the greater number of variations characterizing its parts (or subprocesses)18. For Weiss, the coordination of parts found in the whole was synonymous with a lessening of the variability of the parts since the parts as such were restrained by the emergent stability of the whole. Whereas in a machine the operations of the parts adhered to a centralized or hierarchical control mechanism, in an organic whole the organization emerged rather than being programmed-in. Weiss was also careful to distinguish his understanding of wholes from the holist persuasion that had it that wholes were “disembodied superagencies” which could somehow be extracted from the interactional operations making up organic wholes.
Donna Haraway in her very informative work on the metaphors used by embryologists, particularly metaphors of “crystals”, “fabrics” and “fields,” points out how Weiss’ work on connective tissue and scar tissue formation emphasized “self-organization” which Weiss invoked for what he perceived as the lack of “instructive outside intervention”. He also contended that the self-organizing structures he was observing were emergent, supra-cellular in nature, the latter status comprising their field-like properties16. Indeed, Weiss went even further in his electron microscopy studies of the lamella of amphibian larvae in which he declared the whole system as a mix of many factors providing, “...a singularly suitable object for the study of those organizational factors residing in the body that impose a higher degree of order upon tissue components than that attainable by self-organization.” Indeed, in studying the fibrous structure of these membranes, Weiss concluded, “The fibers, by virtue of their interactions with each other and with their environment, would determine a field of forces with energetically distinguished equilibrium points spaced in the indicated cubic lattice pattern. The pattern of the emergent system of higher order thus would result from the fact that the interacting units themselves have a distinctly nonrandom, patterned constitution” (Weiss quoted in 16:169) Weiss made it a point to emphasize that although students are told of the many marvels of evolution, there was also an equally or even more important conservation of structure at work in living organisms, “a reshuffling, re-sorting and recombination with emergent novelty and progressive improvement–yes; but it all had to start with the full complement of the minimum vital implements to begin with because it’s still the same assortment in the simplest amoeba and the highest metazoan.”
Weiss even avowed that cosmologists need to reconsider the idea that it was all random in the beginning, contending instead that it all started off with some sort of order. Weiss’ research and explanatory strategies can be said to be his own unique spin on Waddington’s delving downward (see 19), but what Weiss called a re-emergence after having first dissected down to “micro-micro-level”.
Autopoiesis, closure, and autonomy
There are at least two ways a self-referential structure could generate and maintain wholeness. First, there is a containment of the system due to the bending back of the causal influence. If the cause leads to an effect which in turn reinforces the cause, then this circular loop acts to unite cause and effect so that outside influences play a secondary role. That is, as a self-referential structure, the system is sequestered from its environment by a boundary established by this very self-referentiality. Second, in a self-referential structure the reciprocity of cause and effect implies they are bound together in a dependency–with emergent wholeness constituted by this state of mutual dependence.
The apotheosis of the self-referential understanding of organic is undoubtedly the theory of autopoiesis put forward by the Chilean biologists Humberto Maturana and the late Francisco Varela20. Maturana had earlier tie-ins to cybernetics with his seminal research on the animal nervous system conducted with several of the original members of the Macy cybernetics conferences where circular causality and feedback had been such central concerns. The term “autopoiesis” comes from both “poiesis,” a Greek term meaning “production,” and “auto-“ for “self,” so that “autopoiesis” means “self-production.” Indeed, there is a correspondence here with the phrase “self-organization” that can be clearly seen in the usual form of the latter term in non-English languages as “auto-organization.”
The term “autopoiesis” appeared for the first time in 1974 as a way for defining, in a manner reminiscent of Kant, what was unique about living beings in contrast to the merely inert: beings ceaselessly producing themselves in a self-referential fashion. Maturana and Varela posited a fundamentally self-referential structure organized in order to maintain the very organization of which it itself was the embodiment. If this sounds like a tautology then the reader is on the right track since, according to the thesis of autopoiesis, living wholes are self-referential in their very essence by being composed of a network of production processes of components which through their interactions regenerate and realize the network which produces them. Thiss self-referential circularity acts to create a self-contained identity, a boundary circumscribed state which Varela termed closure, but which I think is fairer to be called “referential closure.”
In his masterful study of the history of how cognitive science became trapped in a mechanical model, the French social philosopher Jean-Pierre Dupuy points out that although several of the important pioneers in information theory, e.g., Anatol Rapoport, Donald Mackay, and others, had serious qualms about the wholesale assimilation of information theoretic constructs into the study of living organism, it unfortunately became a general assumption that living creatures could be sufficiently understood along the lines of input/output devices not unlike those found in communication channels and computers1.
It was in opposition to such a notion that Weiss offered his fierce critique for not allowing enough autonomy to the organism. Weiss rejected the information theoretic “open” systems stance of early cyberneticians such as McCulloch’s by considering the nervous system as not cut-off from the external world yet not operating according to a simple input/output process. Rather, input acts as a kind of trigger or releasing mechanism which “chooses” among modes of autonomous functionings going in the nervous system complex.
Likewise, Maturana and Varela presented their idea of autopoiesis as a counter to the viewpoint that living organisms were open systems in constant information and energy exchanges with their environment. Varela asked: If living systems were open in such a manner, how they could maintain their integrity in the face of the fact that their cellular parts were constantly perishing? He offered his idea of closure to express what he perceived as a preliminary fact, namely. that environmental inputs first need to be assimilated by the organic whole in its activity as an ongoing integral unity before the organism could then react to them. Maturana and Varela offered an analogy: a pilot flying blind in conditions of very poor visibility relying only on the instrumentation in the cockpit where any environmental inputs are mere blips on the radar screen. This closure operated as a kind of internal autonomy whereby the organism was much more than a mere reactor to input. The central role of referential closure in the theory of autopoiesis was another round in the continuing rejection of mechanistic and reductionist perspectives by emergentist-oriented thinkers.
Realizing that his turn to self-referentiality in order to characterize autopoietic systems lacked credible support in science and mathematics, Varela set out to derive a three valued logic borrowing from the "the algebra of logic" devised by George Spencer Browne which would ground self-referentiality deep in the conceptual foundations of thought and nature21,22. For lucid expositions of this aspect of Varela’s work plus a host of other difficult matters made accessible for further insight see the works of Robin Robertson (many are available at http://www.angelfire.com/super/magicrobin/).
Since self-referentiality was pushed down to a fundamental level, Varela’s approach should more accurately be considered a sort of pan-self-referentialism predicated on the presumption that nature is pervasively self-referential. By making this hylozoist move, referential closure is supposedly no longer so mysterious for after all it’s just an unfolding of what’s already there infolded! But such a move really just serves to shift the conundrum elsewhere as in all hylozoist “resolutions”: for now there is the new riddle of why self-referentiality is not apparent where it’s supposed to be, e.g., in rocks!
The problematic aspect of autopoiesis fell in with Varela’s general avoidance of both morphogenetic change as well as the specific term “emergence.” All that Varela offered concerning the possibility of real change in biological organisms was the vague and diminished idea of some kind of genetic drift since what truly interested Varela was not the coming into existence of emergent wholes but their nature.
Probably because of its highly formalistic manner as well as its reluctance to come to terms with morphogenesis, the theory of autopoiesis has had its biggest impact outside of biology, particularly in post-modern theories of society and culture, e.g., the work of Niklas Luhmann whose work (e.g., 23) has been linked up with post-modernism (primarily on the European continent and in Great Britain). It should also be noted that Varela himself, before his untimely death, went on to play key role in European renditions of artificial life, utilizing the formalisms of the latter to represent the referential closure of neo-emergentist computational phenomena, in this way providing a nice counterbalance to the still prevalent American emphasis on an information theoretic, input/output model in the computational emergence of artificial life.
The original article, published as Swenson, R. (1982). “Autocatakinetics, Yes—Autopoiesis, No: Steps toward a unified theory of evolutionary ordering,” International Journal of General Systems, 21:207-228, is available from here.