home > book reviews > Ludwig von Bertalanffy
Google > Ludwig von Bertalanffy


Books

Short works

Books : reviews

Ludwig von Bertalanffy.
General System Theory: foundations, development, applications: revised edn.
George Braziller. 1968

rating : 2.5 : great stuff
review : 26 October 2009

This book, already 40 years old, is itself a collection of (lightly edited) papers and presentations from the 1940s, 50s and 60s. So my review is necessarily part historical. However, my main interest is in the science. As part of my research in complexity science and emergence, I'm going back and reading some of the precursor literature. It is easy to assume that everything is new and just discovered by the current generation, especially in such a rapidly progressing areas as Computer Science (my students at least seem to think that anything before the mid-1990s is prehistoric). But everything has a history (I can boggle those same students by telling them that object-orientation is over 40 years old; that apparently rapid progress is mostly a feature of hardware changes!), and complexity science has a particularly rich one. Even before I read this collection, I was aware that General System(s) Theory is a significant precursor; now I realise how insightful Bertalanffy was. This is a fascinating mix of the outdated, the now obvious, and the still prescient.

There are rather more quotes from the text than usual for a review, because I want to capture the main points of the argument, and it seems best to do so in the original author's own words.

The first component of Bertalanffy's General System Theory is the system part. GST is a theory of systems, of wholes, rather than a reductionist theory of components, of parts. It grew out of thinking of biological organisms holistically:

p12. The present author, in the early [19]20's ... advocated an organismic conception in biology which emphasizes consideration of the organism as a whole or system, and sees the main objective of biological sciences in the discovery of the principles of organization at its various levels.

The point is that the reductionist view works only if the component parts are weakly coupled, and can be linearly composed into wholes without disrupting or affecting their individual behaviours. This is just not true of non-linear systems where the parts are closely coupled (the very properties that make something a "system", rather than an aggregation), and a new approach is needed.

pp18-19. The system problem is essentially the problem of the limitations of analytical procedures in science. ... "Analytical procedure" means that an entity investigated be resolved into, and hence can be constituted or reconstituted from, the parts put together, ...
     Application of the analytical procedure depends on two conditions. The first is that interactions between "parts" be non-existent or weak enough to be neglected ... The second condition is that the relations describing the behavior of parts be linear ...
     These conditions are not fulfilled in the entities called systems, i.e., consisting of parts "in interaction."

Bertalanffy defines systems as [p38] "sets of elements standing in interrelation". It is not just the components, but their relationships, how they interact, that make up a system. It is this extra feature that gives rise to "non-summative", "emergent", or system-level properties not seen in the individual elements.

pp54-55. In dealing with complexes of "elements," three different kinds of distinction may be made--i.e., 1. according to their number; 2. according to their species; 3. according to the relations of elements. ...
     In cases 1 and 2, the complex may be understood as the … sum of elements considered in isolation. In case 3, not only the elements should be known, but also the relations between them. Characteristics of the first kind may be called summative, of the second kind constitutive. We can also say that summative characteristics of an element are those which are the same within and outside the complex; they may therefore be obtained by means of summation of characteristics and behavior of elements as known in isolation. Constitutive characteristics are those which are dependent on the specific relations within the complex; for understanding such characteristics we therefore must know not only the parts, but also the relations.
     ...
     The meaning of the somewhat mystical expression, "the whole is more than the sum of parts" is simply that constitutive characteristics are not explainable from the characteristics of isolated parts. The characteristics of the complex, therefore, compared to those of the elements, appear as "new" or "emergent." If, however, we know the total of parts contained in a system and the relations between them, the behavior of the system may be derived from the behavior of the parts. We can also say: While we can conceive of a sum as being composed gradually, a system as total of parts with its interrelations has to be conceived of as being composed instantly.

pp67-68. We may define summativity by saying that a complex can be built up, step by step, by putting together the first separate elements; conversely, the characteristics of the complex can be analyzed completely into those of the separate elements. This is true for those complexes which we may call "heaps," ...
     … it is … necessary to emphasize the non-summative character of physical and biological systems because the methodological attitude has been, and is yet to a large extent, determined by the mechanistic program … the "concept of organism" … states, according to Russell, that the laws governing the behavior of the parts can be stated only by considering the place of the parts in the whole. Russell rejects this view. … It is true that the principles of summativity are applicable to the living organism to a certain extent. ... This applies to those phenomena we shall define later as occurring in highly "mechanized" partial systems. But [it] is profoundly untrue with respect exactly to the basic and primary biological phenomena. … the behavior of an element is different within the system from what it is in isolation. You cannot sum up the behavior of the whole from the isolated parts, and you have to take into account the relations between the various subordinated systems and the systems which are super-ordinated to them in order to understand the behavior of the parts. Analysis and artificial isolation are useful, but in no way sufficient, methods of biological experimentation and theory.

The second component of GST is the general part. This idea here is that systems have general properties that can be used to analyse whole classes of systems, in addition to their specific properties that apply only to particular instances of systems. Which are which?

p34. Which principles are common to the several levels of organization and so may legitimately be transferred from one level to another, and which are specific so that transfer leads to dangerous fallacies?

It is important to note that principles can be general, but not to take the analogies too far, to imply some kind of identity of the systems that exhibit common principles.

p118. The fact that simple growth laws apply to social entities such as manufacturing companies, to urbanization, division of labor, etc., proves that in these respects the "organismic analogy" is correct. In spite of the historians' protests, the application of theoretical models, in particular, the model of dynamic, open and adaptive systems … to the historical process certainly makes sense. This does not imply "biologism," i.e., reduction of social to biological concepts, but indicates system principles applying in both fields.

Putting these ideas together gives GST: a non-reductionist theory of systems, of the organisation and interrelations of the parts, with general rules and laws governing the organisational principles.

p49. We come, then, to a conception which in contrast to reductionism, we may call perspectivism. We cannot reduce the biological, behavioral, and social levels to the lowest level, that of the constructs and laws of physics. We can, however, find constructs and possibly laws within the individual levels. ... The unifying principle is that we find organization at all levels.

Not only is GST non-reductionist in that it considers relationships between parts, it is non-static in that it considers the dynamics of those parts and relationships. Systems are not in equilibrium; they change.

p23. Concepts and models of equilibrium, homeostasis, adjustment, etc., are suitable for the maintenance of systems, but inadequate for phenomena of change, differentiation, evolution, negentropy, production of improbable states, creativity, building-up of tensions, self-realization, emergence, etc.; as indeed Cannon realized when he acknowledged, beside homeostasis, a "heterostasis" including phenomena of the latter nature.

So behaviour is important. And because we are talking about systems, we cannot understand the behaviour of the parts in isolation. Their behaviour is fundamentally influenced and controlled by the context: where in the system organisation they are.

p31. It is necessary to study not only parts and processes in isolation, but also to solve the decisive problems found in the organization and order unifying them, resulting from dynamic interaction of parts, and making the behavior of parts different when studied in isolation or within the whole.

This is the main contrast with classical physics, that epitome of the reductionist sciences. The underlying world view of classical physics is not appropriate for non-linear, highly organised, complex systems, and so a new approach is needed. Today we talk of complexity science; Bertalanffy called it General System Theory.

p34-35. Classical physics ... was highly successful in developing the theory of unorganized complexity. ... The theory of unorganized complexity is ultimately rooted in the laws of chance and probability and in the second law of thermodynamics. In contrast, the fundamental problem today is that of organized complexity. Concepts like those of organization, wholeness, directiveness, teleology, and differentiation are alien to conventional physics. However, they pop up everywhere in the biological, behavioral and social sciences ...
... The classical modes of thinking ... fail in the case of interaction of a large but limited number of elements or processes.

Today we might speak of mesoscopic scale: too many elements to treat all individually and exactly, but too few to treat en mass and statistically. However, there is a difference: note Bertalanffy's constant emphasis on the interactions (relationships between elements), not just the elements themselves. Even macroscopic systems can be complex, if they have structured organisation, not mere average properties.

Despite this move away from classical physics, most of the formalisms given in the book are in terms of Ordinary Differential Equations, and how consideration of these can lead to general results.

pp62-63. Mathematically trivial as these examples are, they illustrate … that certain laws of nature can be arrived at … in a purely formal way. The equations discussed signify no more than that the rather general system of equation …, its development into a Taylor series and suitable conditions have been applied. In this sense such laws are "a priori," independent from their physical, chemical, biological, sociological, etc., interpretation. In other words, this shows the existence of a general system theory which deals with formal characteristics of systems, concrete facts appearing as their special applications by defining variables and parameters.

Bertalanffy does allow that ODEs are merely one formalism, and that it might not be adequate for certain kinds of system, particularly ones exhibiting hysteresis (a dependence on history, not just on current state).

pp56-57. [an ODE] definition of "system" is ... by no means general. It abstracts from spatial and temporal conditions, which would be expressed by partial differential equations. It also abstracts from a possible dependence of happenings on the previous history of the system ("hysteresis" in a broad sense); consideration of this would make the system into integro-differential equations ... Introduction of such equations would have a definite meaning: The system under consideration would be not only a spatial but also a temporal whole.

He goes further, and says that there are other (non-DE) areas of mathematics applicable to GST including, cybernetics, game theory, decision theory, and topology (network and graph theory). At the time (and indeed, still today) these other areas are not as exploited as they might be:

p100. information theory has been hailed as a "major breakthrough," but outside the original technological field, contributions have remained scarce. … in biology, DNA is spoken of as "coded information" and of "breaking the code" when the structure of nucleic acids is elucidated, use of the term information is a façon de parler rather than application of information theory in the technical sense … "Information theory, although useful for computer design and network analysis, has so far not found a significant place in biology" (Bell, 1962).
     [BELL, E., "Oogenesis," C. P. Raven, review, Science, 135 (1962), 1056.]

Other definitions of systems include consideration of

p253. UC - set of all couplings between the elements and the elements and environment; ST - set of all states and all transitions between states

Now, thinking computationally, we are very used to "ST", but still have less of a handle on the computational view of "UC". We need a way to capture the dynamics of the relationships and couplings. This is more than just class diagrams with associations: we also need a view of coupling strength, and more importantly, a high level view of the coupling dynamics: changes in relationships and coupling strengths. We need to move the focus from the nodes to the (dynamics of the) edges of the graphs. Indeed, part of the problem is how to model novelty: not just new nodes and links, but new kinds of nodes and links. This would require a system of ODEs that is dynamically changing, with new equations being produced by the system itself. Tricky.

There are many things that haven't changed in the last half-century, in fact. I was actually in the air when reading the following, which caused a wry grin:

p91. Anybody crossing continents by jet with incredible speed and having to spend endless hours waiting, queuing, being herded in airports, can easily realize that the physical techniques in air travel are at their best, while "organizational" techniques still are on a most primitive level.

There is a third component to GST: it is a study of open systems, interacting with their environment. The emphasis is on taking in ("eating") energy and information, rather than on what is given back to the environment ("excreting"), but both are necessary in an open system.

p39. Conventional physics deals only with closed systems, i.e., systems which are considered to be isolated from their environment. ... Thermodynamics expressly declares that its laws apply only to closed systems. In particular, the second principle of thermodynamics states that, in a closed system … eventually the process comes to a stop at a state of equilibrium. … the tendency towards maximum entropy … is the tendency to maximum disorder.
     However, we find systems which by their very nature and definition are not closed systems. Every living organism is essentially an open system. It maintains itself in a continuous inflow and outflow, a building up and breaking down of components, never being, so long as it is alive, in a state of chemical and thermodynamic equilibrium but maintained in a so-called steady state which is distinct from the latter. This is the very essence of that fundamental phenomenon of life which is called metabolism, the chemical processes within living cells. What now? Obviously, the conventional formulations of physics are, in principle, inapplicable to the living organism qua open system and steady state, and we may well suspect that many characteristics of living systems which are paradoxical in view of the laws of physics are a consequence of this fact.

Today we might consider that dynamical systems theory has taken over the mantle of much of GST, with its trajectories, attractors, non-linearity, and general results. Some early ideas can be seen in Bertalanffy's writings:

p254. trajectories may converge towards a stable node represented by the equilibrium point, may approach it as stable focus in damped oscillations, or cycle around it in undamped oscillations (stable solutions); or else diverge from an unstable node, travel away from an unstable focus in oscillations, or from a saddle point (unstable solutions).

Unsurprisingly, there is no mention of strange attractors, since this term was not coined until later. But other dynamical systems concepts are covered (albeit in passing):

p254. A central notion of dynamical theory is that of stability, i.e. the response of a system to perturbation. … stability arguments without actual solution of the differential equations (direct method) and for non-linear systems are possible by introduction of so-called Liapunov functions which are essentially generalized energy functions, the sign of which indicates whether or not an equilibrium is asymptotically stable

But today's dynamical systems theory is still mainly a study of closed systems (in that they have no environmental inputs, although they are often dissipative, and hence essentially losing energy or information). Bertalanffy points out that closed systems have two very different properties from open systems, and that intuitions built up about what is possible in closed systems can lead us very astray when considering open (for example, biological) systems, leading to all kinds of apparently paradoxical results.

The first of these differences concerns final states. Open systems exhibit what Bertalanffy calls "equifinality": they can end in the same place despite different starting points.

p40. In any closed system, the final state is unequivocally determined by the initial conditions... If either the initial conditions or the process is altered, the final state will also be changed. This is not so in open systems. Here, the same final state may be reached from different initial conditions and in different ways. This is what is called equifinality ... It can be shown … that open systems, insofar as they attain a steady state, must show equifinality

(Today, we might talk of "attractors" in dissipative systems.) In fact, the use of final state values in formulae is often forbidden in biology, yet is nothing at all strange or "vitalistic".

p77. The "teleological" final-value formula therefore is only a transformation of the differential equation indicating actual conditions. … physics makes ample use of such final-value formulas because the fact is mathematically clear and nobody attributes an anthropomorphic "foresight" of the goal to a physical system. Biologists, on the other hand, often regarded such formulas as somewhat uncanny, either fearing a hidden vitalism, or else considering such teleology or goal-directedness as "proof" for vitalism. For with respect to animate rather than to inanimate nature, we tend to compare finalistic processes with human foresight of the goal; while, in fact, we are dealing with obvious, even mathematically trivial relations.

The second difference between open and closed systems deals with that old chestnut of the Second Law of Thermodynamics and the increase in entropy.

p41. In open systems ... we have not only production of entropy due to irreversible processes, but also import of entropy which may well be negative. This is the case in the living organism which imports complex molecules high in free energy. Thus, living systems, maintaining themselves in a steady state, can avoid the increase of entropy, and may even develop towards states of increased order and organization.

Bertalanffy notes that some of the relationships in a system can be competitive (eg, sheep and rabbits competing for grass in a Lotka-Volterra system), which can result in extinction of parts of the system. Paradoxically, a predator-prey interaction (eg, foxes and rabbits) does not so readily lead to extinction.

p66. in Volterra's equations ... competition eventually leads to the extermination of the species with the smaller growth capacity; a predator-prey relation only leads to periodic oscillation of the numbers of the species concerned around a mean value. These relations have been stated for biocoenotic systems, but it may well be that they have also sociological implications.

He points out that these ideas of parts of the system competing with other of its parts seems counter-intuitive, but are in fact essential.

p66. If we are speaking of "systems," we mean "wholes" or "unities." Then it seems paradoxical that, with respect to a whole, the concept of competition between its parts is introduced. In fact, however, these apparently contradictory statements both belong to the essentials of systems. Every whole is based upon the competition of its elements

Putting together these ideas of interrelationships, organisation, competition, and dynamics, inspired by developmental biology, leads Bertalanffy to the concept of progressive mechanisation. "Progress" requires differentiation and specialisation. This is achieved by a weakening of interactions, which leads to differentiation and segregation into to a hierarchical structure, which develops into more weakly coupled (less system-like) components. However, this has a corresponding downside: a loss of flexibility.

pp68-69. [where] the interactions between the elements decrease with time ... the system passes from a state of wholeness to a state of independence of the elements. ….
     … the organization of biological wholes is built up by differentiation of an original whole which segregates into parts. ...
     … segregation into subordinate partial systems implies an increase of complexity in the system. Such transition towards higher order presupposes a supply of energy, and energy is delivered continuously into the system only if the latter is an open system, taking energy from its environment. ….
     ... Increasing mechanization means increasing determination of elements to functions only dependent on themselves, and consequent loss of regulability which rests in the system as a whole, owing to the interrelations present. The smaller the interaction coefficients become, … the more "machine-like" is the system--i.e., like a sum of independent parts.
     This ... "progressive mechanization," plays an important role in biology. ... transition from behavior as a whole to summative behavior takes place. ... Mechanization, however, is never complete in the biological realm; even though the organism is partly mechanized, it still remains a unitary system

p70. In this contrast between wholeness and sum lies the tragical tension in any biological, psychological and sociological evolution. Progress is possible only by passing from a state of undifferentiated wholeness to differentiation of parts. This implies, however, that the parts become fixed with respect to a certain action. Therefore progressive segregation also means progressive mechanization. Progressive mechanization, however, implies loss of regulability. As long as a system is a unitary whole, a disturbance will be followed by the attainment of a new stationary state, due to the interactions within the system. The system is self-regulating. If, however, the system is split up into independent causal chains, regulability disappears. The partial processes will go on irrespective of each other. ...
     Progress is possible only by subdivision of an initially unitary action into actions of specialized parts. This, however, means at the same time impoverishment, loss of performances still possible in the undetermined state. The more parts are specialized in a certain way, the more they are irreplaceable, and loss of parts may lead to the breakdown of the total system. To speak Aristotelian language, every evolution, by unfolding some potentiality, nips in the bud many other possibilities. …

This to me is possibly one of the key parts of the theory. It seems to be an aspect that is not covered in anything like this manner by today's complexity science, which tends to think of itself as antithetical to a hierarchical organisational structure, and tends not to look that deeply into the developmental aspects of systems. GST isn't just a theory of systems, but of systems of systems:

p74. Systems are frequently structured in a way so that their individual members again are systems of the next lower level. …
     Such superposition of systems is called hierarchical order. For its individual levels, again the aspects of wholeness and summativity, progressive mechanization, centralization, finality, etc., apply. Such hierarchical structure and combination into systems of ever higher order, is characteristic of reality as a whole and of fundamental importance especially in biology, psychology and sociology.

This system of systems maintains itself by a dynamic process, with different timescales at different levels.

p160. The living organism is a hierarchical order of open systems. What imposes as an enduring structure at a certain level, in fact, is maintained by continuous exchange of components of the next lower level. Thus, the multicellular organism maintains itself in and by the exchange of cells, the cell in the exchange of cell structures, these in the exchange of composing chemical compounds, etc. As a general rule, turnover rates are the faster the smaller the components envisaged

Timescales are important to keep an open system in its steady but dynamic state out of (thermodynamic) equilibrium.

p126. For the maintenance of "dynamic equilibrium," it is necessary that the rates of processes be exactly harmonized. Only in this way is it possible that certain components can be broken down, so liberating usable energy while, on the other hand, import prevents the system from attaining equilibrium. Fast reactions, also in the organism, lead to chemical equilibrium …; slow reactions do not reach equilibrium but are kept in a steady state. Therefore, the condition for the existence of a chemical system in a steady state is a certain slowness of reactions. Momentary reactions, like those between ions, lead to equilibrium in "infinitely short" time. The maintenance of a steady state in the organism is due to the fact that it is composed of complex carbon compounds; these are, on the one hand, rich in energy but chemically inert, so that the maintenance of considerable chemical potential is possible; on the other hand, rapid and regulated release of this amount of energy is performed by enzyme actions, so that a steady state is maintained.

Bertalanffy idea's of progressive mechanisation leads on to some interesting ideas about the emergence of and nature of individuality, and hence identity, and why some systems (like that of the foxes and rabbits) are not individuals (no progressive centralisation).

pp72-73. From the biological viewpoint, we would emphasize progressive mechanization and centralization. The primitive state is that where the behavior of the system results from the interactions of equipotential parts; progressively, subordination under dominant parts takes place. ....
     Thus, similar to progressive mechanization a principle of progressive centralization is found in biology. ... This viewpoint casts light on an important, but not easily definable concept, that of the individual. …
     … strictly speaking, biological individuality does not exist, but only progressive individualization in evolution and development resulting from progressive centralization, certain parts gaining a dominant role and so determining behavior of the whole. Hence the principle of progressive centralization also constitutes progressive individualization. An individual is to be defined as a centered system, this actually being a limiting case approached in development and evolution so that the organism becomes more unified and "indivisible" ...

This structuring principle also solves some of the "pseudoproblems" of how systems work with relationships that are neither completely one-to-one, nor all-to-all: there is intermediate structure that allows many-to-many linkages whilst at the same time having some more important or central than others.

p73-74. Neglect of the principle of progressive mechanization and centralization has frequently led to pseudoproblems, because only the limiting cases of independent and summative elements, or else complete interaction of equivalent elements were recognized, not the biologically important intermediates. ... every gene influences not one single trait but many, and possibly the total organism (polygeny of characteristics and polypheny of genes). ... the genome as a whole produces the organism as a whole, certain genes, however, preeminently determining the direction of development of certain characters ... ... in the function of the nervous system there was apparently the alternative of considering it either as a sum of mechanisms for the individual functions, or else as a homogeneous nervous net. Here, too, the correct conception is that any function ultimately results from interaction of all parts, but that certain parts of the central nervous system influence it decisively and therefore can be denoted as "centers" for that function.

Now, this talk of "wholeness", and of "centres" reminds me of some of Alexander's work, and I wonder if there has been an influence. However, I suspect that this hierarchical idea might not find favour with Alexander; in his essay "A City is not a Tree" he points out that certain systems, like cities, have many more cross-linkages at all levels, and rather have a lattice structure. I think it would be worthwhile to delve into this structural aspect somewhat further.

One other main precursor of complexity science is cybernetics, invented around the same time. Bertalanffy has a lot to say about cybernetics, pointing out the differences between it and GST. One characteristic difference between cybernetics and GST is from the difference between closed and open systems. Bertalanffy describes cybernetics to be of secondary importance.

p44. the feedback scheme … presupposes structural arrangements of the type mentioned. There are, however, many regulations in the living organism which are of essentially different nature, namely, those where the order is effectuated by a dynamic interplay of processes. …. It can be shown that the primary regulations in organic systems … are of the nature of dynamic interaction. They are based upon the fact that the living organism is an open system, maintaining itself in, or approaching a steady state. Superposed are those regulations which we may call secondary, and which are controlled by fixed arrangements, especially of the feedback type. This state of affairs is a consequence of a general principle of organization which may be called progressive mechanization. At first, systems---biological, neurological, psychological or social---are governed by dynamic interaction of their components; later on, fixed arrangements and conditions of constraint are established which render the system and its parts more efficient, but also gradually diminish and eventually abolish its equipotentiality. Thus, dynamics is the broader aspect, since we can always arrive from general system laws to machinelike function by introducing suitable conditions of constraint, but the opposite is not possible.

He classes cybernetic systems as a special class (subset) of all systems.

p161. feedback systems and "homeostatic" control are a significant but special class of self-regulating systems and phenomena of adaptation

In other places he lists the main differences between open systems and cybernetic feedback, where the two approaches seem here to be more on the same level, just capturing different aspects of systems: mainly thermodynamics versus information.

p150. The basis of the open-system model is the dynamic interaction of its components. The basis of the cybernetic model is the feed-back cycle … in which, by way of feedback of information, a desired value … is maintained, a target is reached, etc. The theory of open systems is a generalized kinetics and thermodynamics. Cybernetic theory is based on feedback and information. Both models have, in respective fields, been successfully applied. However, one has to be aware of their differences and limitations.
     The open-system model in kinetic and thermodynamic formulation does not talk about information. … a feedback system is closed thermodynamically and kinetically; it has no metabolism.
     In an open system increase of order and decrease of entropy is thermodynamically possible. … in a closed feedback mechanism information can only decrease, never increase ...
     An open system may "actively" tend toward a state of higher organization, i.e., it may pass from a lower to a higher state of order owing to conditions in the system. A feedback mechanism can "reactively" reach a state of higher organization owing to "learning," i.e., information fed into the system.
     In summary, the feedback model is preeminently applicable to "secondary" regulations, i.e., regulations based on structural arrangements in the wide sense of the word. Since, however, the structures of the organism are maintained in metabolism and exchange of components, "primary" regulations must evolve from the dynamics in an open system. Increasingly, the organism becomes "mechanized" in the course of development; hence later regulations particularly correspond to feedback mechanisms (homeostasis, goal-directed behavior, etc.).

In fact, he explicitly compares open systems and cybernetic systems as two sides (process and structure) of the same systems coin, and calls for a synthesis. This synthesis is still to occur.

p163. Typical feedback or homeostatic phenomena are "open" with respect to incoming information, but "closed" with respect to matter and energy. The concepts of information theory … correspond therefore to "closed" thermodynamics … rather than … open systems. .... The cybernetic scheme permits … clarification of many important phenomena of self-regulation in physiology and lends itself to information-theoretical analysis. The open-system scheme permits kinetic and thermodynamic analysis.
     … dynamics in open systems and feedback mechanisms are two different model concepts, each in its right in its proper sphere. The open-system model is basically nonmechanistic, and transcends not only conventional thermodynamics, but also one-way causality as is basic in conventional physical theory ... The cybernetic approach retains the Cartesian machine model of the organism, unidirectional causality and closed systems; its novelty lies in the introduction of concepts transcending conventional physics, especially those of information theory. Ultimately, the pair is a modern expression of the ancient antithesis of "process" and "structure"; it will eventually have to be resolved dialectically in some new synthesis.

Towards the end, Bertalanffy attempts to show that GST is more widely applicable than just to biology, and applies it to history, and finally, to psychology. Here he is indulging in a justified rant against overly mechanised and behaviourist views of psychology. Viewing the brain not as a passive reactive system, but as a dynamic, self-maintaining system, leads to a very different view.

p209. "…The stimulus … does not cause a process in an otherwise inert system; it only modifies processes in an autonomously active system" (Bertalanffy, 1937, pp. 133 ff ...)
     [BERTALANFFY, LUDWIG von, Das Gefüge des Lebens, Leipzig, Teubner, 1937.]

This view has interesting parallels with theories of immunology that view the immune system as an active dynamic maintenance system that is perturbed by pathogen input, rather than as a passive "guardian" springing into action only when attacked. He applies a systems theory approach to discussing schizophrenic breakdown. However, there is no indication that the problematic symbolic breakdown might be due to physical/chemical problems in the brain material, rather than at a higher symbolic level, so there is no discussion of whether the problem could therefore be treated by drugs rather than psychotherapy.

He has an interesting take on the "free will" problem being a category error pseudoproblem.

p221. Within the framework developed, the problem of free will or determinism also receives a new and definite meaning. It is a pseudo-problem, resulting from confusion of different levels of experience and of epistemology and metaphysics. We experience ourselves as free, for the simple reason that the category of causality is not applied in direct or immediate experience. Causality is a category applied to bring order into objectivated experience reproduced in symbols. Within the latter, we try to explain mental and behavioral phenomena as causally determined and can do so with increasing approximation by taking into account ever more factors of motivation, by refining conceptual models, etc. … causality is not metaphysical necessity, but is one instrument to bring order into experience, and there are other "perspectives" …, of equal or superior standing.

He finally moves on to what might be called cultural relativism: that the symbolic system built up in the brain might be a product of culture as much as anything.

p248. it seems amply demonstrated that the style of thinking is different in the several civilizations even though Whorf's supposition that this is more or less solely due to linguistic factors, is open to criticism.

There is a rather eyebrow-raising 1960s feel about sentences like

p213. the experience of the child, savage, and non-Westerner, though primitive, nevertheless forms an organized universe.

although only a little later he does say

p231-32. There is no intrinsic justification to consider as "true" representation of the world what we take to be "normal" experience (i.e., the experience of the average adult European of the twentieth century) , and to consider all other sorts of experience that are equally vivid, as merely abnormal, fantastic or, at best, a primitive precursor to our "scientific" world picture.

However, he is not subscribing to what might be called "strong relativism". There is a factual basis to reality, but our culture might affect which aspects of it we choose to study.

p237. This, of course, does not mean that the content of mathematics is "true" only within a certain culture. It is a tautological system of hypothetico-deductive nature, and hence any rational being accepting the premises must agree to all its deductions. But which aspects or perspectives are mathematized depends on the cultural context.

And even that choice isn't totally arbitrary. We are physical beings inhabiting a material world, and we have to get it "right" to at least some degree merely in order to survive.

p239-40. As far as direct experience is concerned, the categories of perception as determined by the biophysiological organization of the species concerned cannot be completely "wrong," fortuitous and arbitrary. Rather they must, in a certain way and to a certain extent, correspond to "reality"---whatever this means in a meta-physical sense. Any organism, man included, is not a mere spectator, looking at the world scene and hence free to adopt spectacles, however distorting, .... Rather he is a reactor and actor in the drama. The organism has to react to stimuli coming from outside .... There is a latitude in what is picked up as a stimulus .... However, its perception must allow the animal to find its way in the world. This would be impossible if the categories of experience, such as space, time, substance, causality, were entirely deceptive. The categories of experience have arisen in biological evolution, and have continually to justify themselves in the struggle for existence. If they would not, in some way, correspond to reality, appropriate reaction would be impossible, and such organism would quickly be eliminated by selection.

So, this is a fascinating account of General Systems Theory as it stood 40-odd years ago. There are deep insights, and lots of food for thought. So, what happened? Why is it not now mainstream? And if it didn't succeed, why might its descendent, complexity science, fare better?

I can only speculate. But it may be that GST was ahead of its time. Reductionism hadn't been mined out: there was still an awful lot of progress to be made, particularly in biology, the "old" reductionist way: genes to sequence, data to mine. Now we have a huge heap of components, it is being made clear much more forcefully that this is insufficient: all the King's horses and all the King's men are having a devil of a time putting Humpty together again. And maybe the mathematics wasn't up to it. We still don't really have a unified theory of material information processing, of open systems that process both matter and information.

Why might we do better now? If we are to do better, part of the answer has to be: "computers". All that complexity just can't be handled analytically. And all the emphasis is on behaviour, on dynamics, and computers are great at exploring dynamics. Maybe this time around, we can make more progress. But to do so effectively, we must remember our history, what came before, so that we don't waste effort reinventing the wheel. We need to stand on the shoulders of giants, and Bertalanffy seems to have a good set of shoulders on him.