Jesper Hoffmeyer
On the Origin of Agency and Life
Cybernetics and Human Knowing 
Vol. 5, no. 1, 1998   pg. 33

Abstract:  Von Foerster has suggested the Möbius strip as a topological representation of the kind of logic pertaining to self- referential cybernetic systems. The Mobius strip offers the conceptual categories of an outside interior and an inside exterior. It is suggested that these categories are realized in natural cybernetic systems through semiotic loops integrating self-reference and other-reference. Autopoiesis and semiosis are supplementary categories. Living systems may be seen as consisting essentially of surfaces inside other surfaces. 

The closure of a membrane around some autocatalytic chemical reaction system is an attractive candidate for a first step towards the origin of a living system. 

A spheric surface defines an inside-outside asymmetry and opens the possibility for communicative activity across the membrane. If some modest kind of co-operation arose in populations of closed surfaces these surfaces might become interfaces for real communication. 

Two further steps would be needed for these surfaces to become true anticipatory biological systems. The surface and its internal autocatalytic system would have to produce a written (digital) record of ist own components, and the surface would have to devise means for controlling the translation process whereby components are produced. 

Only in this way can the surface become a temporal being, an autonomous agent capable of making distinctions and engaging itself in future-oriented internal or external modification. Such a system has been termed a code-dual system.

A Theory of the Subject

Gertrudis van der Vijver in her recent interview with Heinz von Foerster observed that one of the criticisms on second order cybernetics, which has led to what is now vaguely called third order cybernetics, is "that second order cybernetics tried to develop a theory of the observer but did not have a theory of the subject"(Vijver 1997: 5). 

Replying to this Von Foerster states the need for "an epistemological salto mortale, because the moment you open your mouth you open the mouth, but to identify therefore what is coming out of the mouth which has been opened is reflecting the open mouth", and he points to the Mobius strip as an interesting topological representation of the logic of such self-reference. 

The strange thing about the Mobius strip is that inside and outside are co-extensive so that there is only one side, and yet the Mobius strip does in fact produce an asymmetry between interior and exterior. The Mobius strip thus introduces the paradoxical conceptual categories of an "inside exterior" and an "outside interior". But how would such conceptual categories be reflected in real systems, i.e. real living subjects? Answering this question may be of help if we are to construct a theory of the origin of "subject-ness".   

The French philosopher Maurice Merleau-Ponty saw subjects as temporal beings characterized by a strange form of non- coincidence (Merleau-Ponty 1945). Monika Langer sums up Merleau-Ponty's standpoint thus: 

"Our reflections on time are themselves taking place within time, our reflections on subjectivity are themselves a part of our subjectivity. We can never become one (coincide) withourselves precisely because of this non-coincidence gap" (Langer 1989: 129).   
I have suggested that even non-human living systems might be said to be subjects in a very general sense of the word: "Living creatures are self-referential, they have a history, they react selectively to their surroundings and they participate in the evolutionary incorporation of the present in the future." (Hoffmeyer 1996a:51). 

The key to this kind of temporal non-coincidence is the property typical to all life which Claus Emmeche and I termed "Code-duality" (Hoffmeyer and Emmeche 1991). 


Code-duality refers to the fact that living systems always form a unity of two coded and interacting messages, the analogly coded message of the organism itself and its re-description in the digital code of DNA. 

As analog codes the organisms recognise and interact with each other in the ecological space giving rise to a horizontal semiotic system (the ecological hierarchy of Stanley Salthe (1985)), while as digital codes they are passively carried forward in time between generations (after eventual recombination through meiosis and fertilisation in sexually reproducing species). 

This of course is the process responsible for nature's vertical semiotic system, the genealogical hierarchy (i.e. genegenotype/ phenotype - deme- monophyletic lineage, cf. Salthe, 1985). Seen in this light heredity may be described as 'semiotic survival' (Hoffmeyer, 1997c). 

The dynamics of Code-duality assures a perpetual retranslating back and forth between two differently coded versions of the living system creating a branching flow of semi-mutable messages down through generations. The term message refers to the possibility for meaningful interpretation inside the contextual system of the lineage and its eco-semiotic niche.   

Code-dual systems are anticipatory in the sense, that the digital code (the genotype pool) records specifications which did work well enough in the past, and which are then used by the analogly coded organisms to cope with the immediate future, thereby eventually assuring the semiotic survival into the more distant future. 

This of course is anticipation in the most primitive sense of extrapolation from the past. But the fundamentally semiotic character of this system very early in evolution assured the creation of sophisticated sense facilities to strengthen anticipation.

Code-duality is a unique feature of living systems and I have suggested it as a defining criterium for being alive (Hoffmeyer 1996a). This criterium would exclude computers since these have not (at least not yet) been constructed to depend on the creative activity of an analogly coded version interacting with realworld processes in such a way as to test the fitness of the digital specifications necessary for its own construction. 

I shall suggest in this paper that one other key to a theory of the origin of life or subject-ness is a process of asymmetry-formation through membrane closing followed by the development of mechanisms for semiotic interaction across the membrane (Hoffmeyer 1998). As we shall see it is the semiotic dimension which accounts for the paradoxical categories of the "inside-exterior" (the Umwelt) and the "outside-interior" ("the semiotic niche" (Hoffmeyer 1996a))   Much as the concept of autopoiesis has helped in establishing the understandingof the self-asserting capacity of cybernetic systems it may nevertheless have tended to overshadow the semiotic aspect of life. 

Self-production and self-reference is only half the story, the other half is other-reference or representation (and, in fact, other- production cf. e.g. Griffith and Gray 1994). A single genome and a single organism is just one instant in a grand historical process stretching billions of years back in time, and nothing in the cell or the organism makes sense if not seen in the perspective of the organizing influence of this deeply semiotic process. 

Insides and Outsides

To ask for the origin of life is to ask for the origin of the environment. Living organisms are inscribed in their environments much like patterns woven into a carpet. The two cannot get apart and yet there seems to be a distinct asymmetry in their relation. 

From the point of view of the organism the environment is everything outside of it, while from the point of view of the environment organisms are encapsulated packages having appeared inside it like an infection. And in saying so we can perhaps glimpse at an even deeper but also more questionable asymmetry. 

For while it feels rather obvious to say that the organism has a point of view, even though the vlsual metaphor is misleading, it doesn't feel natural to ascribe a point of view to the environment. The environment is there for the organism, not vice versa. Is this asymmetry justified?   I have come to see this as a very crucial question. 

'A point of view' is a very strange thing indeed if seen from the point of view of science. The title of Thomas Nagel's famous book from 1986, The View from Nowhere, captures well what I take to be the essence of the scientific endeavour throughout the last few hundred years. 

What Nagel criticises in his book is not the objectivistic strategy as such, but the belief that reality is in a narrow sense identical to objective reality: 'The fundamental idea behind the objective impulse is that the world is not our world.This idea can be betrayed if we turn objective comprehensibility into a new standard of reality. That is an error because the fact that reality extends beyond what is available to our original perspective does not mean that all of it is available to some transcendent perspective that we can reach from here' (Nagel 1986: 18).   Nagel's position however leaves us with the problem of how to deal scientifically with the eventual non-objectivistic aspect of nature, i.e. that nature is replete with (living) systems having points of views. 

In other words, how do we address the problem of the legitimacy of the organism-environment asymmetry in a satisfactory way?   For a start let us observe that what is outside actually always tend to be inside something else, and what is outside at one instant of time may be turned inside at a later time, as is for instance the case during early embryogeny in mammals when an invagination in the spherical stage of the blastula gives rise to the internally layered stage of the gastrula.   

Yet at a more fundamental level insides are always defined by the boundary surrounding them, and in living systems such a boundary (e.g. the cell membrane or the skin) is itself a physical structure of a certain thickness and internal structure. 

From the point of view of the membrane or the skin even the 'inside' ofthe organism will in principle be an 'outside'. And this may well be the key to a solution to the organism-environment asymmetry problem. 

Life is built on a fundamental asymmetry, but this is not an asymmetry between organism and environment. Instead it is an asymmetry produced by any closed membrane (e.g. the skin) which separates the world into two equally excluded parts: an internal part and an external part. 

The membranes of living systems - at whatever level, i.e. whether they encircle sub-cellular organelles, cells, tissues, organs, or organisms - are in fact best described as interfaces facilitating a highly regulated exchange of signs between interiors and exteriors. 

Life should fundamentally be seen as organized around the nested set of  membranes or interfaces which we call organisms (Hoffmeyer 1998).

Agency and Situated Time

Bruno Latour has discussed "the heroic effort” we have invested in "eliminating from the mind, from the production of science, from the ontogenetic development, from the history of science, and... from the history of life itself, any trace of history, of time-producing practice" (Latour, 1997). Following a distinction first proposed by the philosopher Isabelle Stengers, Latour suggest the concept of time as virtual: 

"We never encounter time and space, but a multiplicity of interactions with the actants having their own timing, spacing, goals, means and ends" (Latour,1998). 

Time, space and actor (or agent) are thus to be seen as mutually linked aspects of an event or a process, i.e. time is situated and loaded with agency - or, with the term Bruno Latour suggests "intensity".   

Stanley Salthe has defined agency in the following way: "A system has agency with respect to some process if its effects in the world leave a trace of its individuality. Agents cause historical events that contribute to the individuation of other systems" (Salthe, 1993 p.159). 

Agency in this sense is not an exclusive property of humans or even living beings. For instance a snow storm may exhibit agency as it apparently did when, as Salthe mentions, "Jane Byrne became mayor of Chicago because of one snow storm." While we may be allowed to take for granted that the snow storm did not anticipate this result or have any intentions to bring it about it nevertheless did leave its track in the web of American history. And the snow storm itself, where did it come from? Amplification of fluctuations? Causa sui?

Work on robotics has given important inspiration to our thinking about  "autonomous agents" in general. In a fascinating analysis Horst Hendriks-Jansen  has this to say: 

"What constitutes an adequate description of an autonomous agent?.. Whether the aim of the aim of the exercise is engineering or explanation, adequate descriptions can emerge only from interactions between autonomous agents and their environments. The natural kinds of a science of behaviour are emergent activity patterns" (Hendriks-Jansen, 1996, p.l70). 

He introduced the concept interactive emergence to characterise the phenomenon that patterns of activity whose high-level structure cannot be reduced to specific sequences of movements may emerge from the interactions between simple reflexes and the particular environment to which they are adapted" (Hendriks-Jansen 1996,p. 8). And he underlines that the natural kinds of this type of explanation come into existence only as a result of the creature’s situated activity. 

They are not entities that can be reduced to events inside the creature’s head; neither can they be defined by classifying its environment. Interactive situated behaviour cannot be explainedin terms of a deductive or generative law. 

It requires a historical explanation because there can be no rules to predict the sorts of behaviour that might emerge." Situated activity presupposes a kind of fitting between the system and its environment. This poses no problem as long as we talk about robots. But how dowe account for the situatedness of organisms? Hendriks-Jansen contents himself by pointing to natural selection as the explanation for interactive emergence in the organic realm. But this will clearly not do because natural selection already presupposes competing individuals, which again presupposes that organisms insome way "strive" to survive. But how do we explain the interactive emergence of this "striving"?   

Mitchel Resnick, who developed the StarLogo turtles, offers an essential cue to the solution to this problem: "StarLogo turtles come equipped with "senses". They can detect (and distinguish) other turtles nearby, and they can 'sniff' scents in theworld. There is even a built-in primitive to make turtles 'follow a gradient' of ascent - that is, to make turtles turn in the direction where the scent is strongest. Such turtle-turtle and turtle-world interactions are essential for creating and experimenting with self-organizing phenomena. Parallelism is not enough. If each turtle just acts on its own, without any interactions, interesting colony-levelbehaviours will never arise" (Resnick, 1994, p.33). In other words, sensing is of essence here: for entities to interact in patterns of emergence they must posses semiotic competence. 

Interactive emergence then comes close to the dynamics which I have discussed as semetic interaction (of Greek semeion = sign, and etos = habit, Hoffmeyer, 1995; 1997a). 

Semetic interactions refers to interactions in which regularities (habits) developed by one species (or individual, tissue, cell) successively become used (interpreted as signs by individuals of the same or another species, thereby eliciting new habits in this species eventually to become - sooner or later - signs for other individuals, and so on in a branching and unending web integrating the ecosystem of the planet into a global semiosphere (Hoffmeyer, 1996a, Lotman, 1990). 

The semiosphere is a sphere like the atmosphere, the hydrosphere or the biosphere. It penetrates these spheres and consists in signification and communication: sounds, odours, movements, colours, electric fields, waves of any kind, chemical signals, touch,etc.   

The question of how to account for the existence of natural autonomous agents then leads us to the question of the origin of semiotic competence, which accordingto the American linguist and semiotician Thomas A Sebeok is probably congruent with the question of the origin of life: A full understanding of the dynamics of semiosis may in the last analysis turn out to be no less than the definition of life" (Sebeok, 1979, p.26). According to Koichiro Matsuno and Stanley Salthe however semiosic agency is an intimate aspect of material process in our universe and biosemiosis rather must be seen as a highly specified elaboration of a muchmore general principle (Matsuno and Salthe, 1995).

Five Necessary Steps on the Road to the Origin of Life

The standard view of the origin of life has it that the dawning chemistry of life was organised around RNA molecules. But as pointed out not the least by the American' complexity scientist Stuart Kauffman the RNA scenario suffers from serious  problems (Kauffman, 1995). Kauffman not only observes that the chemical  preconditions for the RNA-scenario are far from obvious but also points to more  theoretical considerations contradicting the idea of simple RNA-life. The simplest  existing free-living cells, so-called pleuromona, are highly simplified exemplars of  bacteria containing an estimated number of genes of a few hundred to about a  thousand. Viruses, which are vastly simpler than pleuromona, are not free-living. They are parasites that invade cells, co-opt the cell's metabolic machinery to  accomplish their own self-replication, escape the host cell, and invade another.  

"All free-living cells have at least the minimum diversity of pleuromona" writes  Kauffman and continues: Your antenna should quiver a bit here: Why is there this  minimal complexity? Why can’t a system simpler than pleuromona be alive?" (p. 42).   The answer Kauffman gives to these questions is based on his work with mathematical modelling of "combinatorial chemistry" which shows that when a large enough number of reactions are catalysed in a chemical reaction system, a vast web of catalysed reactions will suddenly crystallise. 

"Such a web, it turns out, is almost certainly autocatalytic - almost certainly self-sustaining, alive" (p.58). Complexity thus is a prerequisite to autocatalytic closure, which again is a prerequisite to life. And Kauffman confidently concludes that: "The secret of life, the wellspring of reproduction, is not to be found in the beauty of Watson-crickpairing, but in the achievement of collective catalytic closure. The roots are deeper than the double helix and are based in chemistry itself. So, in another sense, life - complex, whole, emergent - is simple after all, a natural outgrowth of the world in which we live" (p. 48).   

Much as Stuart Kauffman's work is a brilliant and highly convincing tour deforce turning upside down many of our received ideas on the origin of life it nonetheless leaves a feeling that something important is missing. Clearly at least semiotic competence cannot be accounted for in a satisfactory way without  capacity for making distinctions (Hoffmeyer 1997b). The missing element in Kauffman's scheme of things, we shall suggest, is the idea of the surface as an ontologically primary entity. a natural kind, i.e the idea that life is fundamentally based on surfaces inside surfaces. The achievement of collective catalytic closure is a necessary first step for agency to be realised, but at least four moresteps are needed.

ORIGIN OF LIFE:   Five necessary steps     

1 Autocatalytic closure (Kauffman)     
2 Inside-outside asymmetry (closed surface)     
3 Proto-communication (a community of surfaces)     
4 Digital redescription (code-duality)    . 
5 Formation of an interface (inside-outside loops)

Consider a membrane closing itself around a complex mixture of chemical reactants. Allow for a restricted flow of small molecules across the membrane and suppose that by chance the internal mix of reactants combined with the flow across the membrane produced an autocatalytic system. From a biochemical point of view this all makes good sense (Weber, 1998a, 1998b). Such a system would have a surface and therefore it would posses a fundamental asymmetry between inside andoutside.   

Suppose furthermore that many such surfaces formed a local environment for reciprocal and, more or less, self-sustaining chemical interaction. One might perhaps not yet call this communication because no true actors are involved so far. But the situation seems ripe for semiosis to arise (proto-communication: Step 3 infigure 1)   

Let us now add one further thing to this system: An internal "written self-record". Kauffman himself introduces this expression and defines it as a "record "kept at 0 degrees Kelvin for delayed use (Kauffman, 1996: 91). Now, for good reasons cells did not use 0 degree Kelvin but instead they used something nearly as good: DNA. DNA is a very inert molecule which is in fact protected against the vicissitudes of life by its own lack of reactivity at physiological relevant temperatures. A DNA record is conservative and has a capacity for "memory".   

When such closed surfaces became stabilised due to re-description of their own components in a code of DNA, they might eventually survive even if fatally disturbed, since they could survive not only by 
simple autocatalysis but also by reconstruction based on a translational kind of process (cf. Depew and Weber). And in due time some surfaces might learn to split themselves into two more or less identical systems (Step 4 in figure 1).   

So far a system has been established which - seen from the observers point of view - has an obvious interest in maintaining the needed flow of chemicals across the surface. But the system still has no way to assist the fulfilment of its own "interest", it has no mechanism for goal-oriented modification or action. Thus the system is not an agent in its own interests. It doesn't matter to the system whether it can distinguish features of its environment, and therefore it has not yet any capacity for making distinctions.   

What is needed in addition to the DNA-record is the formation of a feedback link between DNA and environment, so that the surface will have different proteins at its disposal in different situations (Weber et al. 1989). The surface in otherwords must turn into an interface linking the interior and the exterior (step 5 in figure 1). Only then does the system's understanding of its environment matter to the system, and this is how the logic of the Mobius strip becomes realized IN ACTU: relevant parts of the environment becomes internalized as an "inside exterior", a phenomenal world or perceptual model which was called the Umwelt by Jakob von Uexkull (Uexkull 1982[1940], and in the same time the interior becomes externalized as an "outside interior" in the form of "the semiotic niche", i.e. the diffuse segment of the semiosphere which the lineage has learned to master in order to control organismal survival in the semiosphere. And by the same token.his is of course the decisive step in the evolutionary process of attaining true semiotic competence, i.e. the competence to make distinctions in space-time whereformerly there were only differences. 

The semiotic looping of organism and environment into each other through the activity of their interface, the closed membrane, also lies at the root of the strange future-directedness or "intentionality" of life, its "striving" towards growth and multiplication (Hoffmeyer 1996b). The spatial asymmetry between the "inside interior" and the "outside exterior" is coupled to the time asymmetry implicit in the self-referential mechanism of DNAre-description followed by cell division.   

To see this let us finally consider why the redescriptive system needed? Why isn't it enough for the surface to have an internal autocatalytic closure? The answer is that self-ness presupposes temporality, a self must have an internal temporal link for otherwise it would be meaningless to say that the world matters to it. If something should matter to a system then the system must have an existence in time. The "written record" or DNA-description serves as a "present" memory (a proto-value) linking past and future around it. The temporal surface is linked to the spatial surface, the two asymmetries are integrated: time is situated and loaded with agency - or, with the term Bruno Latour suggested "intensity".

Evolution: A Search Through Context Space

Organisms may be seen as living in open context spaces, where the context space can be defined as the total set of N elements exhibiting K potential mutul perceptional links. Only a very diminutive and species specific subset of context space is actually perceived by each single organism. We can call this subset thre Umwelt-space. Evolution can now be seen as an on-going parallel search through context space for regularities worth incorporating into Umwelt-space. Umwelt-space therefore represents a set of species specific expectations of the lineage or individual concerning the range of traits in the environment which will be of importance in the future. The Umwelt-space of a moth, to take an example, will contain no sound emitted by birds or other vertebrate animals with the single exception of the 20.000 hertz frequency used by hunting bats. For good reasons the moth species has learned to anticipate the danger associated to this frequency. 
This anticipation of course is not a mental anticipation, but is bound into the semiotically organised architecture of the senso-motoric system of the moth. The Umwelt-space maps onto a "response space" defined by the set of possible activities of the organism and the mapping relation between Umwelt-space and response space is itself open to evolutionary modulation.  

We shall suggest that in the course of evolution the Umwelt-space of species has tended to increase in sophistication so that ever more complicated patterns of context space could be dealt with. 

Concurrently with this trend there gradually appeared a trend towards the delegation of anticipatory competence from the level of the species to the level of the individual allowing for the evolution of individual learning processes. 

And finally an animal evolved to anticipate its own death and therefore to understand its fundamental alienation from that strange world which will be there even though the animal itself would disappear drawing its Umwelt along into disappearence. The non-coincidence of 1. and 3. person singularis, the human subject, was born.

HOME        BOE       SAL      TEXTE