2 Origins

... the simulation rapidly provided the results our intuition had led us to expect: the spontaneous emergence in this artificial bi-dimensional world of units which self-distinguished by means of the formation of a `membrane', and which showed a capacity of self-repair. ... It is important to mention this article [46] here because it was the first publication on the idea of autopoiesis in English for an international public, which led the international community to take charge of the idea. In addition it anticipated what twenty years later would become the explosive field now called artificial life and cellular automata.

--Varela [45, p. 414]

As recounted by Varela, the concrete idea of autopoiesis arose only slowly, and in particular circumstances of time and place. The development seems to have originated around 1967 [45]. Varela was then a student (or ``apprentice'') under Humberto Maturana in the Department of Sciences at the University of Chile. Maturana was already well known, with an international reputation for his work on the neurophysiology of vision [see e.g., 13]; but he had started to develop a basic dissatisfaction with the idea of ``information'' as being key to understanding brain and cognition.

The problems here are profound, and are still by no means resolved. Nervous systems somehow contrive to imbue mere ``information'' with meaning and significance. Despite the continuing dominance of the ``information processing'' paradigm of cognition, such signification cannot be intrinsic to the raw information. Indeed, the very basis of the Shannon ``information theory'' is, precisely, to separate information from meaning. So signification is something that is somehow generated through the autonomous, self-referring, dynamics of a nervous system, constrained by its embedding in, and coupling with, the world it comes to know.

These ideas of autonomous circular organisation in nervous systems were developed by Maturana over the following several years. While Varela's personal trajectory took him to the USA at this time, the two continued to meet and collaborate together, and with Heinz von Foerster, in discussing these ideas. This culminated in the publication of Maturana's paper Neurophysiology of Cognition [22]. Varela retrospectively identified this as a critical step (to which, as Maturana acknowledged, both Varela and von Foerster had contributed)--``... the indisputable beginning of a turn in a new direction'' [45, p. 412]. In particular, it made the first connection between these issues of self-reference in nervous systems and corresponding, but even more basic, problems regarding the molecular constitution of the biological cell--which is to say, of life itself.

Varela returned to Chile in 1970, now as a colleague of Maturana in the Biology Department at the University of Chile. They deliberately focussed on exploring the nature of the organization of the living organism. At this point the problem was becoming clear as that of understanding or characterising the demarcation between living and non-living systems--but to do so in a way which was abstracted away from the specific biochemical contingencies of any particular form of life. So we see already here exactly the distinction between ``life as it is'' and ``life as it could be'' that eventually provided the intellectual foundation for the field of Artificial Life [17].

By mid-1971, the word itself, autopoiesis, had been coined, and Varela and Maturana had outlined a minimal model or exemplar. By this was meant an imaginary, abstract, chemical world, incorporating just those species and reactions that would suffice to allow the constitution of a minimal autopoietic system. This would therefore serve to isolate, in a very precise and concrete way, just what was, and was not, being pointed at by the new concept.

Thus, the minimal model served an important expository purpose--to give concrete form to what is otherwise a rather difficult and unfamiliar idea; but to do so in a way that strips away the many complex distractions that any real bio-chemical system would additionally present. Moreover, then, as now, no real living cell is yet described in sufficient detail to be able to definitively identify its constitution as autopoietic in any case.

Yet, the minimal model also served another, potentially much more important purpose: as a test of what had now become, not just a concept, but a theory of autopoiesis. Maturana and Varela conjectured that autopoietic organisation was capable of giving rise to characteristically biological phenomenology--in particular, to that most primitive cellular phenomenon of ``self-repair'', of the conservation of composite, macroscopic, cellular organisation even as the microscopic constituents of the cell are continuously turned over.

Ideally, of course, such a test case might have been implemented in ``real'', ``wet'', chemistry. But this is both technically challenging and, in any case, runs the risk of re-introducing many distracting complications that are not relevant to the restricted question being asked. So, instead, they hit on the idea of using a computer simulation; and, in particular, the idea of a low-level, fine-grained (``individual-'' or ``agent-based'') simulation.

This methodological idea was undoubtedly already ``in the air''. Varela himself described it as an ``obvious step'' [45, p. 414]. It drew explicitly on the pioneering work of John von Neumann--work carried out in the early 1950's, but which had been properly published only a few years prior to Varela and Maturana's work [47,29]; and on the seminal development of the Game of Life by Conway et. al., reported just as Varela and Maturana were already working along these lines [12].

On the other hand, some distinctive and highly original innovations were also required. The basic idea of a discrete, two-dimensional, space, with local interactions, clearly derives from von Neumann. However, Varela and Maturana introduced the concept of directly implementing (quasi-)stable ``particles'' which can move across the lattice, rather than relying on the relatively complex and fragile emergence of patterns of underlying node-states with this property (e.g., in the manner of gliders in the Game of Life). They also incorporated a deliberately stochastic dynamics, and inter-particle transformations or reactions--essential to anything motivated by the idea of realising an abstract ``chemistry'' (as opposed to an abstract ``physics''). From a contemporary perspective, writing in the pages of the Artificial Life Journal, these are now essentially standard techniques; but it is important to recognise the originality and insight that they represented in that much earlier era.

In any case, the assistance of Ricardo Uribe of the School of Engineering at the University of Chile was enlisted, and, as reported in the opening quote of this section, a computer simulation of the minimal model was created which ``rapidly'' confirmed the qualitative expectations of Varela and Maturana.

By the end of 1971 then, the theory of autopoiesis in its most basic form--as a theory of the molecular organisation of living cells--had been quite fully articulated, and tested in the form of a highly simplified (and therefore highly demanding) computer simulation. Two texts had been written: a comprehensive technical account, and a draft of a shorter summary article (including the simulation results).

Publication proved difficult. The extended version was rejected by at least five publishers and journals; the short paper was submitted to several journals, including Science and Nature, with similar results. However, following a visit to Chile by Heinz von Foerster in mid-1973, the paper was further revised and submitted to BioSystems (for which von Foerster was a member of the editorial board), and eventually published in mid-1974 [46]. Much more comprehensive publications have, of course, followed [24,23,44]; but as the first English language publication of the theory of autopoiesis to an international audience, the 1974 BioSystems paper remains a key reference, which continues to be widely read and cited. Further, as the first application of agent-based computer modelling to demarcating the living from the non-living state, it occupies a seminal position in the pre-history of what we now call Artificial Life.

Barry.McMullin@dcu.ie