To explain the emergence of teleodynamics, we must also resist the temptation to attribute it to similarly mysterious and counterintuitive phenomena, such as the strange and spooky properties exhibited by quantum-level processes, which similarly appear to violate our familiar notions of cause and effect. Although these might in some cases superficially resemble ententional relationships, such as those involving measurement, action at a distance, or entanglement, these must be excluded for three reasons. First, if we were to make use of any such extraordinary physical phenomena, we would at best only have subst.i.tuted one mystery for another. And even so, this would not explain the difference between non-living and living systems in this respect. As we earlier noted with respect to panpsychic or pansemiotic a.s.sumptions, it would still require us to come up with an explanation for why living and mental processes are a.s.sociated with such radically different causal dynamics than their non-living counterparts. Of course, were someone to show that it was in principle impossible to explain these higher-order properties and processes using the tools of known macroscopic physics and chemistry, that might force us to consider the possible role of these very different quantum properties. But this brings us to the second reason to avoid such an appeal. If we can explain this shift in causal organization without invoking strange causality, it will render the justification for appealing to strange quantum properties irrelevant.

Lastly, although quantum phenomena often seem strange, they are not strange in the sense of exhibiting the kind of causality that organisms and minds do. No one describes any of the strange quantum properties, such as the Casimir effect or quantum entanglement, as consequence-organized behavior. More important, the scale at which we do unambiguously recognize ententional properties is vastly larger than the scale of quantum events, and in between there are only thermodynamic and chemical processes. The basic physical and chemical forces that dominate at our level of scale, from cells to ecosystems, appear to be the only place we find the emergence of the ententional properties-exhibited in living and mental and communicative processes. In short, ententional phenomena are not strange in the way that quantum processes can be strange, and they are manifest only at a much higher level of scale.

With these caveats in mind, we can make a first pa.s.s at identifying the minimal requirements for the emergence of teleodynamic processes; then we will describe a molecular thought experiment that embodies these requirements. Providing a realistic, empirically testable, constructive account of the emergence of teleodynamics in a simple molecular system is only the barest starting point of a theory of the origins of life. It is a long way from explaining even the features exhibited by the simplest known bacteria. It is only a minimal proof of principle. Many more details (some to be explored in subsequent chapters) will need to be filled in to even approach a methodology sufficient to begin to account for the evolution of the higher and more complex forms of teleodynamic processes, such as are found in living organisms and mental processes.

So, in this first step out of mere physics, we will only offer evidence that a spontaneous transition from non-ententional to ententional forms of causal organization is possible. The goal is solely to develop a minimally complicated, thoroughly described, physically plausible model system exemplifying this transition. With such a model system at our disposal, we can begin to ground our accounts of at least simple ententional processes and relationships in more precise physical, chemical, and thermodynamic details. The basic principles developed in this simple system can then later be applied at higher levels of a.n.a.lysis, where we encounter ententional phenomena of considerably more subtle and complex form. To the extent that we develop such a plausible and non-mysterious bridge a.n.a.lysis, it will open the door to a methodology for examining how these higher-level ententional phenomena might have emerged and evolved from the lower-level ententional phenomena exhibited by the first precursors to life.

To conceive of the first lifelike process, we must remind ourselves that it was not reproduced, it had no parent, and therefore it did not evolve. It emerged. It was fitted to its environment by chance alone, not by virtue of having been tested by natural selection, and yet the very fact of its emergence indicates a consistency with its environment. Indeed, it would be circular to argue that the process of natural selection itself evolved by natural selection, or that the reproductive process arose by being reproduced. On the other hand, it is quite likely that the first ancestral counterpart to today's genetically mediated process of reproduction was quite simple and inefficient, and that the first ancestral counterpart to natural selection was considerably less effective at achieving and maintaining favorable variations within a lineage.

With ultimate simplicity, we can expect only a very crude first exemplar of that which modern organisms accomplish with comparatively flawless elegance. In comparison to this first form, even the genetically undisciplined and promiscuous nature of bacterial and viral evolution would appear like the essence of clockwork precision. This is because the process we now see enshrined in these mechanisms of reproduction is the outcome of billions of years of fine-tuning by natural selection, and the entire web of contemporary Earth life descends from the one winner in a billion-year compet.i.tion among possible alternatives. This contemporary near uniformity is almost certainly the outcome of an older period of disorderly inefficient experiments in adaptation and inheritance. The tricks that life currently employs to generate its many lineages and diverse adaptations likely evolved to be ever more evolvable over these many billions of years. How these critical and powerful semiotic and functional capacities arose will be the subject of a later chapter. But the processes that honed the evolutionary process itself cannot have been any form of evolution as we now know it, and as we will shortly see, the very logic of evolution must have also emerged in this transition.

The challenge of this book is to explore the possibility that teleodynamic processes can emerge spontaneously by natural laws from physical and molecular processes devoid of these properties. But we should be wary of looking to contemporary biology for clues. The debates that raged in past centuries over the plausibility of spontaneous generation suggest a useful caution. Surrounded by teeming life, so imperceptibly minuscule as to require extreme care to avoid contamination, spontaneous generation theorists were fooled again and again into thinking they had demonstrated the emergence of life from non-life. Like the early Enlightenment biologists immersed in a world teeming with invisible life, we are immersed in a world of ubiquitous ententional a.s.sumptions that we barely recognize as potential contaminants. Trying to explain this emergent transition, from the perspective of hindsight, and within our currently well-honed biological context, risks a.n.a.logous contamination. Setting our thought experiments in a teleologically barren context is thus the equivalent of testing spontaneous generation in a sterile environment. This is not as easy as it sounds.

AUTOCATALYSIS.

One of the most significant insights concerning the nature of life has come from studies of thermodynamic processes that are far from equilibrium, which I have collectively referred to as morphodynamic processes. In organisms, far-from-equilibrium conditions are maintained by the complex cycles of molecular reactions const.i.tuting cell metabolism. Normally, in a closed system, each instance of a given chemical reaction will slightly decrease the future probability of similar reactions, due to the depletion of precursors, the increase of products, the dissipation of the energy of the reaction, and how all these changes affect the relative probability of the reaction running forwards versus backwards. So, in general, a closed molecular system will tend toward some steady state with fewer and fewer asymmetric reactions occurring over time, as the overall distribution of reactions runs in directions that offset each other-the second law of thermodynamics at work in chemistry. In contrast, in chemical systems maintained far from equilibrium, where the conditions for asymmetric reaction probabilities are not reduced, non-equilibrium dynamics can produce some striking features. The most relevant are morphodynamic chemical processes.

One of the most relevant cla.s.ses of non-equilibrium chemical process is autocatalysis. It is relevant both because there are many a.n.a.logues to this in living cell metabolism and because it can arise spontaneously as a transient, locally deviant, non-equilibrium process. A catalyst is a molecule that, because of its allosteric geometry and energetic characteristics, increases the probability of some other chemical reaction taking place without itself being altered in the process. It thus introduces a thermodynamic bias into a chemical reaction as a consequence of its shape with respect to other molecules. Autocatalysis is a special case of catalytic chemical reactions in which a small set of catalysts each augment the production of another member of the set, so that ultimately all members of the set are produced. This has the effect of producing a runaway increase in the molecules of the autocatalytic set at the expense of other molecular forms, until all substrates are exhausted. Autocatalysis is thus briefly a self-amplifying chemical process that proceeds at ever-higher rates, producing more of the same catalysts with every iteration of the reaction.

Because of the widespread presence of chemical cycles in living systems, autocatalysis offers a suggestive starting point for discussing the origins of life, and it has been promoted by many as the chemical a.n.a.logue of self-replication. Unfortunately, it requires the remarkably coincidental coalescence of reciprocally interlocking molecules, making it a fairly rare and spontaneously improbable chemical condition. But is its spontaneous occurrence sufficiently probable to serve as a one-in-a-billion-years springboard for life?

The theoretical biologist Stuart Kauffman has, however, argued that this intuition about the improbability of spontaneous autocatalysis is mistaken.2 Using principles from graph theory to model the networks of reaction of catalytic interactions, he argues that autocatalysis will be almost inevitable under not too extreme chemical conditions. Because catalysis is a more-or-less effect, based on only the relative precision of "fitting" between different molecules, a polymeric "soup" with sufficient diversity of molecular forms will tend to include many molecules able to at least weakly catalyze reactions among others in the soup. This plurality of potential catalytic possibilities can be envisioned as comprising a sort of network of relationships between molecular types (forming the nodes in the network), and their synthesis and breakdown relationships (const.i.tuting branching relationships in the network). As molecular diversity increases (or with high average catalytic potential between any two molecules), the effective connectivity of the corresponding network will increase as well. With increasing network connectivity, re-entrant circles of reactions become more common. So, as molecular shape diversity increases, the probability of spontaneous circles of catalysis will increase as well. We may then expect that multiple interlinked autocatalytic loops will be present in sufficiently "rich" polymeric soups that are not so concentrated as to form glue, and that this provides for a sort of initial exploration for the most robust and sustainable variants. But although autocatalysis is a function of molecular diversity, it turns around and generates molecular h.o.m.ogeneity, since molecules that tend to produce other molecules from a linked autocatalytic set rapidly replace other forms that do not.

At the point where the molecular diversity in the solution reaches the stage where circles of reactions of this sort start to become common there is a discontinuity in the dynamics of chemical reactions. With the introduction of autocatalysis, there is a break in the symmetry of the distribution of probabilities of the different types of chemical reactions. Members of this autocatalytic constellation will rapidly increase in proportion to all other molecular types, and as they do so, this very process will speed up too. It is a.n.a.logous to the rapid multiplication and growth of new organisms introduced into a pristine environment. But also like over-reproduction, it will be a transient dynamic that quickly undermines its base.

A higher-order special case of autocatalysis was investigated by the physical chemist Manfred Eigen in which two or more autocatalytic cycles are interdependently linked to create cycles of cycles.3 These were dubbed hypercycles. In general, a hypercycle is a second-order autocatalytic cycle formed when a product of one autocatalytic cycle is an essential element in another autocatalytic cycle, which produces a product that is essential to the first; thus forming an autocatalytic cycle of autocatalytic cycles. There are probably many ways this kind of higher-order autocatalysis can be const.i.tuted, but the obvious cases involve autocatalytic cycles contributing substrates to one another, or autocatalytic cycles partially overlapping by sharing catalysts in common. Both occur in living cells, where higher-order networks of linked catalytic reactions produce highly complex and interwoven metabolic and synthetic cycles of cell biochemistry that are mutually reinforcing.

Of course, all catalysis-and especially autocatalysis, because of its runaway dynamic-is constrained by the availability of substrate molecules and energy. This is even more critical for hypercycles, because they include multiple autocatalytic cycles, each of which is subject to such extrinsic dependencies, as well as on the continual productivity of all the other linked cycles. When the raw materials are used up, or energy for molecular reactions is no longer available, reactions stop and the set of interdependent catalysts dissipates. Autocatalysis is thus self-promoting, but not self-regulating or self-preserving. And hypercycles will be even more sensitive to substrate conditions and dissipation, because they are subject to weakest-link effects. Substrate limitations on any one subcycle of a hypercycle will limit all, and the slowest catalytic process will exert a rate-limiting influence on all other linked subcycles.

In addition to these problems, there is an intrinsic limitation on the initiation and persistence of any autocatalytic process. Such factors as the relative concentrations of the interdependent catalysts with respect to any other molecules not involved in autocatalysis, their tendency to react with other molecules in ways that are non-conducive to autocatalysis, and the ubiquitous effects of spontaneous diffusion, all will limit autocatalysis. These are not unlikely limiting conditions, even in laboratory settings, and they pose significant impediments to spontaneous autocatalysis, especially if spontaneously achieving autocatalysis among a random a.s.sortment of molecules depends on significant molecular diversity. Moreover, these difficulties increase geometrically with the complexity of the autocatalytic network of reactions, making even modest catalytic complexity highly unlikely. Nevertheless, complex reciprocally catalytic networks of molecular interactions characterize the metabolic processes of living cells, so it is far from impossible.

CONTAINMENT.

The second ubiquitous feature of life is containment. All units of life on Earth, whether single cells, multicelled organisms, or viruses, are self-contained. All cells have lipid cell membranes that prevent indiscriminant diffusion of molecules between inside and outside. Even viruses, which lack other sorts of molecular systems, are encased by protein sh.e.l.ls or more complex coverings that both protect their nucleic acid payload and aid in getting the virus insinuated into a host cell. Self-containment is a ubiquitous feature of life because life depends on the structural contiguity of its molecular systems remaining intact and unchanging over long periods of time. Containment is the most obvious expression of the importance of constraint for the successful persistence of life.

Containment creates physical individuality. A boundary that distinguishes inside and outside is almost synonymous with the self/other distinction, both functionally and metaphorically. The characteristic unit of contained individuality in life is a cell, and although there is individuation at higher levels in multicellular organisms (e.g., organism bodies), and at lower levels with respect to cellular subdivisions (e.g., organelles), it is the cellular level of life that exemplifies the most robust and omnipresent unit of functional individuality throughout the living world. The cell membrane is a boundary distinguishing a continuously maintained self-similar milieu inside from a varying and unconstrained outside world. Though neither impermeable nor inert, its role is vested in what cannot happen as a result. The constraint on molecular movements and interactions that containment provides is a necessary const.i.tutive factor in all living systems.

Finiteness of the contained materials also makes the state s.p.a.ce of their interrelationships manageable. Insulation from extrinsic factors that are too many, too distributed, and too unpredictable likewise is an essential contributor to the maintenance of self-similarity across time. And simply maintaining material proximity of potentially diffusible components that must interact regularly and predictably to maintain living processes is an important constraint on the increase in entropy. For all these reasons, and because of its ubiquity in the living world, many scenarios concerned with the origins of life treat containment as primary.

But containment is a double-edged sword. Maintaining components in continual proximity and excluding new interactions increases similarity and predictability from moment to moment, but this happens for the same reason that a closed thermodynamic system quickly and inevitably runs down to a maximum entropy state. Complete closure rapidly leads to stasis. Self-similarity rather than stasis is the basis for life's individual units, however. Partial or periodic containment, on the other hand, is able to contribute to both. Without spatial constraint on the correlated loci of interdependent molecules and processes, neither life nor evolution seems possible.

Interestingly, most abstract conceptions of Darwinian processes ignore this factor. In simulations, for example, self-replication is almost universally represented without any consideration of containment, but this may again reflect more on the abstraction process than on the constraint requirement. A-Life, evolutionary programming, genetic algorithms, and social evolution theories may not need to deal with the problem of containment because the symbolic medium provides it implicitly in another form. Conventionalization of sign vehicles allows the selective identification of each representative unit by distinctive characteristics. Von Neumann's replicating mechanism can also ignore containment because the physical mechanism is identical to its container. It is presumed to be a permanently interconnected device that by design coheres as a physical machine without diffusion tendencies. Still, the measurement of what must be replicated at least requires the a.s.sessment of essential boundaries to determine what must be "contained" in the representation, as "self" (to be copied) versus other. But where we cannot take for granted the self-identification or intrinsic structural coherence of all essential functional elements, some coherence-maintaining, proximity-maintaining mechanism must be present. For any molecular system whose elements are necessarily available for chemical interaction, there must be some additional barrier to diffusion and interaction consisting of relatively non-reactive molecules that limits this potential. Whether in the form of extrinsically maintained proximity and coherence or intrinsically integrated binding mechanisms, these functional roles must somehow be fulfilled to sustain a functional unity.

Containment of organisms and viruses is generally achieved by a molecular process called self-a.s.sembly. This is a ubiquitous feature of all organisms and viruses, and is responsible for more than mere containment. Virtually all multimolecular complexes that const.i.tute the molecular machines doing the molecular work of the cell spontaneously a.s.semble. So their structure is not explicitly constructed in the way engineered devices or fabricated artifacts must be. As noted in previous chapters, organisms and machines differ critically in this respect. A machine must have its parts chosen, gathered together, aligned, interconnected, and set in motion by extrinsic processes. The process of a.s.sembly of molecular building blocks into large multicomponent structures in living cells, in contrast, seldom requires the existence of an external means of accomplishing this feat. These structures tend to spontaneously self-a.s.semble, and in effect "fall" together. This makes life radically unlike the self-reproducing device that von Neumann envisioned. Recall that his self-reproducing device had to include both instruction-copying mechanisms and a machine-fabrication mechanism, with the latter being capable of fabricating a copy of itself from the instructions. The engineering paradigm takes as given the need to impose structure from without; but this is not an intrinsic requirement, and it is almost entirely irrelevant for life. In many ways, this is more than just an advantage that life has over engineering. It is a necessity. This critical const.i.tutive feature helps to explain a good deal about what makes life unusual, and makes evolution possible.

Spontaneous self-a.s.sembly is neither as esoteric nor as special as the term implies. One does not need to imagine parts that are animated and seeking each other out or the existence of special intrinsic a.s.semblers. Self-a.s.sembly of macromolecular structures is essentially a special case of crystallization. Like crystal lattice formation, the growth of multi-unit macromolecular structures is an expression of the intrinsic geometry of component molecules, the collective symmetries these offer in aggregate, and the lower energy state of the crystallized forms. It is in this sense an expression of a thermodynamic orthograde tendency to reduce total entropy. It's just that molecular shape and charge features, and their relationships to other molecules-including especially to the surrounding water-are responsible for the spontaneity of their growth. The accretion of molecules into crystallinelike arrangements is thus a function of thermodynamics, but the influence of their structural and charge characteristics may contribute to the amplification of constraints, thereby generating regularities in the ways they form into aggregates; a morphodynamic consequence.

The combination of these features ultimately determines which among the possible macroscopic arrangements is most likely. Macromolecular structures within cells result from the same symmetry logic that produces the geometrically regular growth of crystal lattices. Individual molecules can align with each other with respect to complementary structural symmetries in their shapes and charges so long as there are nearly complementary symmetries in these molecular features. Crystal-like growth of cellular structures, composed of hundreds or thousands of identical or iterated complementary elements, is common and plays a central role in cellular-scale architecture.

Paradigm examples of self-a.s.sembling macromolecular structures in cells include a wide array of laminar and tubular structures within cells. These include various forms of cell membrane that are found both inside and outside cells, and tubular structures which provide the internal three-dimensional support for cells, the thoroughfares along which molecules and vesicles are moved, the capacity for differential mobility and reshaping of cell geometry as in pseudopodium extension, and the structural elements that form the core of flagella and cilia.

Consider one of the most ubiquitous of these forms: microtubules. Most microtubule formation is based on the spontaneous, spirally symmetric packing of component molecules (tubulin) to produce crystal prism shapes that can be indefinitely extended. Spontaneous precipitation onto the end of a tubule occurs when there is a sufficient concentration of component tubulin molecules and an appropriate ionic concentration for their (orthograde) precipitation. Extension and collapse of these tubular structures at specific loci within a cell can thus be regulated by a variation in substrate availability, solute factors, and other relevant conditions. This coupling of spontaneous formation and conditionality is the key to their functional usefulness. Note that the structure of these tubes is not a function of genetically specified protein structure alone, because the genetic code only incompletely constrains its range of possible three-dimensional structure (which also depends on contextual conditions), and even this structure may be consistent with many different semi-regular packing configurations. Ultimately, both the specification of the structure and the symmetries that drive the self-a.s.sembly process are context-derived-emergent from laws of symmetry and thermodynamics. Much of the structural information and construction work thus comes "for free," so to speak, in the sense that it is neither maintained by natural selection nor strongly determined by genetic information. The genetic information might rather be thought of as maintaining all these variables within constraints that make such spontaneous tendencies highly likely.

Another paradigmatic cla.s.s of self-a.s.sembling structures are the sheets and surface structures that const.i.tute the walls and part.i.tions of cellular architecture, including lipid bilayers, protein and carbohydrate matrices, and the protein capsid containers of viruses. These also tend to form spontaneously via the symmetric aggregation of large numbers of identical components. Probably the simplest const.i.tute the lipid bilayer sheets that form the external and internal membranes of all cells. These form as a result of the polar hydrophilic-hydrophobic structure of lipid molecules. In an aqueous solution, the hydrophobic "tails" of lipid molecules tend to aggregate together, exposing only the hydrophilic ends to the surrounding aqueous solution. Aggregate lipid b.a.l.l.s and bubbles are energetically more "relaxed" than individual lipid molecules in water because aggregation allows neighboring molecules' hydrophobic tails to hide each other from water molecules. Again, this is an expression of the interaction between thermodynamic and structural features of the molecules, which bias how they interact as a consequence of congregating due to settling into a lower energy state. But it is also possible for lipids to form sheets, either at the interface of water and a non-aqueous medium, or with a complementary sheet of lipid molecules, resulting in the formation of a bilayer sheet in which the hydrophobic tails of each lipid molecule faces inward from both sides. This is again a more relaxed geometry than forming irregular aggregates in water; and so free lipid molecules will spontaneously "prefer" to aggregate into sheets when conditions are favorable to this, and these sheets will tend to form into closed spheres if they grow large enough. Because of the highly specific symmetry of the molecular orientation in a sheet, and the structural instability of growing sheets only a couple of molecules thick, lipid bilayers form more effectively and predictably when there is some surface to provide an appropriate planar template to bias this orientation and growth.

Other molecular types are also capable of forming into surfaces with slightly different properties. For example, the formation of viral sh.e.l.ls is more a.n.a.logous to crystal growth than lipid sheet formation, and yet is also quite distinctive from both in that it often involves the formation of regular polyhedral structures. The capsid units that form these sh.e.l.ls are typically composed of proteins. Individual protein molecules or protein multimers fold into shapes that tend to bind edge-to-edge with one another due to structural symmetries. Proteins that a.s.sume regular polygonal or polyhedral shapes will tend to aggregate spontaneously into tessellated sheets, also due to the lower energy of the close packing enabled by this geometric symmetry. Polyhedral prisms created by tessellated surfaces of protein molecules will be the most stable of these forms. This is because all molecules in a three-dimensional polyhedron are bound to one another and surrounded on many sides in a low-energy symmetrical configuration that is thereby structurally resistant to external forces. The regular or semi-regular polyhedrons thus formed typically enclose a DNA- or RNA-filled core.

FIGURE 10.1: Two self-organizing molecular processes common to all life: autocatalysis and self-a.s.sembly. Left: in autocatalysis, one molecule catalyzes a reaction that produces a second catalyst as a byproduct, which in turn catalyzes the first. In this depiction, energy is released by the breaking of bonds of split substrate molecules and causes the process to be self-sustaining so long as substrate molecules are present. Right: three different self-a.s.sembling molecular process are depicted: self-a.s.sembly of the protein and RNA components of a virus; self-a.s.sembly of a lipid bilayer due to hydrophobic and hydrophilic affinities of these molecules' polar structure (hydrophilic "tails" are forced together); and self-a.s.sembly of tubulin molecules into a microtubule, one of the major components forming the flagella that propel bacteria and other mobile cells. Details of each of these processes are described in the text.

After incorporation into a host cell, viral genes are released from this core and repeatedly transcribed by host cell mechanisms to generate high concentrations of capsule proteins. At the same time, viral genes are being replicated in high numbers as well. The spontaneous formation of hundreds of viral sh.e.l.ls in the context of hundreds of gene replicas thus has a high probability of encapsulating the genes that produce them, even without additional packaging mechanisms (though packaging is typically aided by other molecules synthesized from viral genes). The spontaneous self-a.s.sembly of viral capsule molecules into regular containers allows viruses to be elegantly simple and minimalistic, and consequently highly efficient replicating systems.

SYNERGY.

Probably the most prescient and abstract characterization of the dynamic logic of organism design was provided by the philosopher Immanuel Kant. In a 1790 critique of the problem of teleology in nature, he argued that "An organized being is then not a mere machine, for that has merely motive power, but it possesses in itself formative power of a self-propagating kind which it communicates to its materials though they have it not of themselves" (italics in the original).4 Implicit in Kant's abstract characterization of "formative power" is the fact that organisms are organized so as to resist dissolution by replacing and repairing their degraded components and structural characteristics, and eventually replacing themselves altogether by reproduction. More important, as described in the epigraph to this chapter, he emphasizes that this is a reciprocal process. No component process is prior to any other. Kant's characterization is prescient in another way that is relevant to our enterprise. In this essay, he is puzzling over the question of whether there is something like intrinsic teleology in organisms. Kant concludes that this formative reciprocity const.i.tutes what he calls "intrinsic finality." Although modern accounts can be far more concrete and explicit than Kant's, by virtue of their incorporation of over two hundred years of biological science, this knowledge can also be a source of distraction. In these intervening centuries we have of course discovered a vast hidden world of molecules, chemical processes, and cellular interactions, as well as forms of life and half-life (e.g., viruses) that no one in Kant's time could have dreamed of. But it is not clear that contemporary definitions of life are actually any more fundamental and general. Only able to reason about life in the abstract, Kant focused on life's distinctive dynamical organization, and so it is the synergy of living processes that stands out for him. Today, it is possible to add flesh to Kant's skeletal definition and in so doing demonstrate its prescience.

As we have seen, both autocatalysis and self-a.s.sembly are morphodynamic molecular processes that are capable of occurring spontaneously in a wide variety of conditions. What they share in common is a dependency on molecular shape-effects and a propensity for promoting rapid self-amplifying regularities. They are both non-linear processes, like compound interest, and are as a result intrinsically constraint-amplifying. What is amplified in each case is a bias based on symmetry and shape complementarities between molecules, which together produce increasing self-similarity (of molecular types or geometric regularity, respectively) over time. These processes are achieved and sustained by a constant availability of raw materials and energy in the case of catalysis, or local concentrations of identical molecules in the case of self-a.s.sembly.

These requirements, however, make autocatalytic cycles (and especially hypercycles) intrinsically susceptible to diffusion and side reactions, and make self-a.s.sembly a self-limiting process due to its intrinsically concentration-depleting dynamic.

Thus, despite the spontaneous potential for formation of autocatalytic sets, autocatalysis is a fleeting and transient occurrence in the non-living world because it is self-undermining and self-limiting. Precisely because of its deviation-amplifying dynamic, autocatalysis will rapidly deplete the local environment of required substrates. As the process continues and catalysis slows with declining concentration, the diffusion tendencies of the second law of thermodynamics will no longer be compensated by replacement of newly synthesized molecules, and the component catalysts will tend to diffuse away from available substrates and away from other members of the set with which they could interact, if substrate molecules were available.

Similarly, self-a.s.sembly is dependent on a local concentration of free-floating component molecules that will tend to give up their kinetic energy to settle into a lower-energy state within a growing sheet, matrix, or tube. Changes in concentrations of components, changes in the rest of the chemical milieu, or the presence or absence of different external structural biases can halt growth, or even trigger the shift to alternative binding symmetry, disrupting large-scale structure growth. And growth itself depletes this local substrate concentration. Without the continual availability of unbound substrates, growth ceases. Growth must also outpace forces that tend to disrupt the coherence of the enlarging-and therefore more fragile-structure. As sheets expand, or molecules link up into larger surfaces or longer tubes, they may decrease in their overall energetic stability to the point where further growth becomes limited by structural fragility. For this reason, sheets tend to form into stable closed configurations (e.g., polyhedrons or tubes) that are both more robust to external perturbation and nearly as energetically relaxed. As closed structures, however, they may cease to be able to grow further.

But these intrinsic limitations of autocatalysis and self-a.s.sembly processes are also a source of potential synergy. The conditions produced by each of these processes and their limitations together comprise a complementary and reciprocally supportive effect.

Self-a.s.sembly provides the conditions that are most critical for sustaining autocatalysis: the proximity of reciprocally interdependent catalysts. The major consequence of self-a.s.sembling containment is local blockage of molecular diffusion. Spatial proximity of all const.i.tuents of an autocatalytic set is an essential necessary precondition for sustained autocatalysis, but in solution, spontaneous diffusion will undermine this requirement. Containment can, however, maintain proximity irrespective of whether catalysis is or is not taking place, and even in the absence of substrates.

And reciprocally, autocatalysis complements self-a.s.sembly. The major consequence of autocatalysis is the continual production of identical molecules in the same region, whereas self-a.s.sembly is most robust if the concentration of component molecules is maintained despite depletion due to this process. So self-a.s.sembly is most reliable in conditions where there is continual replenishment of component structural molecules up to the point of structural closure, at which point additional components are irrelevant. Thus rapid production of identical molecules at an accelerating pace by autocatalysis is specifically consistent with the conditions promoting self-a.s.sembly.

The reciprocal complementarity of these self-organizing processes means that spontaneous linkage of autocatalysis with self-a.s.sembly containment is a possibility. This would occur in the case that an autocatalytic cycle produced a byproduct molecule that itself was conducive to spontaneous self-a.s.sembly into a shape that could act as a container. In this configuration, each self-organizing process would reciprocally contribute to conditions promoting the stability, persistence, or recurrence of the other. Raw materials would be produced that are conducive to container formation. An enclosing structure would tend to form in the vicinity of the most active autocatalysis and thus would spontaneously tend to enclose autocatalytic set molecules within it. And closure of the container would prevent dissipation of the components of the autocatalytic set. Although this containment would also eventually halt autocatalysis by limiting availability of substrates, it would do so only after closure.

AUTOGENS.

This complementarity of morphodynamic processes can produce far more than merely the creation of a kind of molecular bottle. Their reciprocity produces a special kind of emergent stability, unavailable to either process in isolation. Although continuous catalysis is prevented in the enclosed state, it remains potentiated by the proximity of all the essential molecules of the autocatalytic set. In the event that such a container were to be broken up or breached by agitation or chemical disruption in the vicinity of new substrate molecules, autocatalysis would recommence and new container-forming molecules would be synthesized. This could lead to either re-formation of the original or formation of two or more new containers of catalysts in its place. In other words, the reciprocal complementarity of these two self-organizing processes creates the potential for self-repair, self-reconst.i.tution, and even self-replication in a minimal form.

FIGURE 10.2: Two forms of simple autogenic molecular processes. Left: the formation of polyhedral capsules which contain catalysts that reciprocally catalyze the synthesis of each other and also produce molecules that tend to spontaneously self-a.s.semble into these polyhedral capsules thereby likely to enclose the catalysts that generate them. Right: the formation of a tubular form of encapsulation, which although not fully closed will tend to restrict movement of contained catalysts along its length, but will tend to be increasingly susceptible to partial breakage and release of reciprocal catalysts as it grows longer. Both will tend to re-form or replicate additional copies if disrupted in the presence of appropriate catalytic substrates.

Importantly, this is not a property that is likely restricted to just a tiny set of molecular forms. It is in effect a generic cla.s.s of chemical dynamics that may be achievable in numerous quite diverse ways. In general terms, the key requirements are only a reciprocal coupling of a spontaneous component production process and a spontaneous proximity maintenance process that encompa.s.ses all essential components. Though these reciprocal relationships are modestly restrictive, they are not extreme, and their spontaneous occurrence is made more probable because they are likely realizable in quite diverse chemical environments.

Candidate molecular conditions will be discussed in a later chapter, but the co-facilitation between two such morphodynamic processes could likely be realized by a wide variety of molecular substrates, and in the laboratory it could probably even be achieved using non-organic materials. Although the stereochemical and energetic requirements for such a system to form spontaneously are nevertheless quite limiting, similar limitations apply to any set of molecular interactions proposed as precursors for living processes. It is the relative simplicity of this system of dynamical relationships, and the diversity of ways this co-facilitation could be achieved, that makes this type of molecular complex a far more likely candidate for a first spontaneous self-reproducing system. With respect to known processes extracted from life, protein-based autocatalytic sets enclosed by the a.n.a.logues of viral capsules could likely be synthesized in the laboratory. So this is also a realistically testable hypothesis.

Because of the familiarity of such component cla.s.ses of molecular processes, and despite the currently hypothetical nature of the process, I think it is justified to a.s.sume that the characteristics I have attributed to these simple molecular systems are realistic extrapolations, which justify considering the implications for emergent dynamic effects. Indeed, I offer this as the potentially simplest cla.s.s of teleodynamic systems. Because of its simplicity and the non-mysterious nature of the properties that such a molecular process would exhibit, if we can demonstrate that its dynamical properties have resulted in a clear, higher-order orthograde dynamic than its component processes with unambiguous ententional properties, even if these are minimal, then we have at our disposal a model system capable of fully demystifying this realm without reducing it to mere thermodynamics or sneaking homuncular properties into the account.

FIGURE 10.3: A cartoon depiction of various stages of polyhedral autogenic structures (as in Figure 10.2) self-a.s.sembling. The final image (e) shows the breakup and rea.s.sembly due to collision.

Elsewhere, I have called this sort of hypothetical molecular system an autocell.5 Unfortunately, I have found this term to be somewhat limiting and misleading, since the components described are not necessarily cellular, and the term is not mnemonically descriptive of its most distinctive properties. For the remainder of the book, I will adopt the more descriptive term autogen for the whole cla.s.s of related minimal teleodynamical systems. This term captures what is perhaps its most distinctive defining feature: being a self-generating system. In this respect, it is closely related to Maturana and Varela's autopoiesis, though referring to a distinct dynamical unit process rather than a process more generally, for which I have reserved the more general term telodynamic. The term autogen is also easily modified to apply to a broader cla.s.s of related forms by describing any form of self-encapsulating, self-repairing, self-replicating system that is const.i.tuted by reciprocal morphodynamic processes as autogenic, and describing the process, appropriately, as autogenesis.

Autogenesis is not, however, meant to be descriptive of just any process of self-generation. It is conceivable that a looser interpretation of this term could refer for example to replication of identical molecules in autocatalytic processes or to the hypothetical self-replication of RNA molecules.6 The term is reserved for simple dynamical systems that accomplish self-generation by virtue of harnessing the co-dependent reciprocity of component morphodynamic processes. Though I am here provisionally a.s.suming that this is only a property exhibited by simple molecular systems, there is no theoretical reason to a.s.sume that it is impossible in radically different forms. Experimentation with simple molecular autogenic systems will ultimately be required to determine the parameters for formation, persistence, and replication of molecular autogens, as well as to determine the supportive properties required of substrate molecules and the surrounding molecular environment. But the logical and theoretical plausibility of this form of dynamical organization provides sufficient justification to consider its implications for the emergence of ententional phenomena. So, in advance of definitive laboratory creation of autogenic molecular processes, and irrespective of whether this property can be realized in other kinds of substrates, many reliable extrapolations concerning the properties of such systems can be explored; and in particular, those that exemplify teleodynamic organization.

Autogenic theory is superficially similar to Manfred Eigen's hypercycle theory to the extent that each of these two self-promoting processes also promotes the other in some way, forming the a.n.a.logue of a causal circle of causal circles, so to speak. But the resemblance to hypercycle architecture stops there, and in other respects autogenic theory is fundamentally different. As we saw above, an autocatalytic cycle is susceptible to self-undermining and self-limiting dynamics, and a hypercycle is doubly (or multiply) susceptible to this (depending on the number of substrate-dependent subcycles that const.i.tute it). Each is a potential weak link that if broken will be catastrophic for the larger synergy. In contrast, the reciprocal linkage of the two complementary morphodynamic processes const.i.tuting an autogen has the opposite effect. Though each component process is self-undermining in isolation and co-dependent, together they are reciprocally self-limiting, so that their self-undermining features are reciprocally counteracted. Thus, whereas substrate exhaustion leads to both autocatalytic and hypercycle cessation and component dispersion, an autogenic system will establish its capacity to re-form before exhausting substrates, so long as closure is completed before this point is reached. Each process contributes essential boundary conditions for the other, providing the equivalent of a continual supportive environment for both processes. One might then describe an autogen as a hierarchic hypercyclic system, with each self-organizing component acting as supporting environment or context for the other.

To the extent that autogenesis provides the possibility for self-reconst.i.tution after partial disruption, it also provides a potential mechanism for self-reproduction. By the same process that enables an autogen to form in the first place, fractional components of a disrupted autogen-including sh.e.l.l and catalytic molecules in close proximity and new substrate molecules from the surrounding medium-will be able to reconst.i.tute a new complete whole. In other words, a disrupted autogen will be as likely to produce two identical autogens as one. So an autogen can accomplish in molecular terms-and with considerably more compactness than previously envisioned-what von Neumann demanded of self-reproduction: it can reproduce itself as well as its physical capacity to reproduce itself. Remarkably, it can accomplish this without many of the attributes normally a.s.sumed to be essential for life: for example, molecular template-based replication of components and of the template molecules, incessant far-from-equilibrium thermodynamics, semi-permeable membrane containment, and so on.

This self-reconst.i.tuting dynamics provides an active self-similarity-maintaining quality which const.i.tutes a form of individuality, or "self," that does not otherwise exist outside of living processes. There is both an individual autogenic ident.i.ty, as a closed, inert, but potentially self-reconst.i.tuting unit, and a self-maintaining lineage ident.i.ty, due to the transmission of relatively invariant intrinsic dynamical constraints and molecular types from generation to generation as a result of replication. Self-reconst.i.tution does not completely maintain material ident.i.ty across time because it allows for molecular replacement, and it does not maintain energetic or dynamical continuity across time either, since it may persist in a static phase for extended periods. But this self-reconst.i.tution capacity does maintain a persistent and distinctive locus of dynamical organization that maintains self-similarity across time and changing conditions. And yet ultimately there is no material continuity, as autogens are disrupted only to be reconst.i.tuted and replicated with newly synthesized components. Only the continuity of the constraints that determine the autogenic causal architecture is maintained across repeated iterations of dissolution and reconst.i.tution.

So an autogen has an ident.i.ty only with respect to this persistent general pattern of constraint maintenance and replication, and irrespective of any particular molecular const.i.tuents. Indeed, it is the continuity of the inheritance of constraints on its molecular dynamics that const.i.tutes this individuality. But ident.i.ty may vary as an autogen lineage evolves variant forms of this defining dynamics (see the section on autogenic evolution below). For all these reasons, an autogen self and autogen lineage ident.i.ty are examples of efficacious general types, in a philosophical sense (see chapter 6). To be more specific, an autogen is an empirical type determined only by the continuity of these dynamical constraints, which are themselves expressions of dynamical limitation-potential modes of change not expressed.

Autogenic organization only exists with respect to a relevant supportive environment. So autogenic individuation is also only defined with respect to a particular type of environment. Ident.i.ty and environment are thus reciprocally defined and determined with respect to each other, because the same molecular configuration in a non-supportive environment lacks any of the defining properties of autogenesis. Indeed, the very possibility for autogen existence can be described as one of the possible micro configurations of a certain cla.s.s of environments with the molecular const.i.tution conducive to autogen formation.

This is the ultimate basis for what Jacob von Uexkull called an Umwelt: the organism-relevant external world.7 This critical autogen-environment relationship is, however, curiously at once a sufficient but not a constant necessary condition for autogenesis to persist. The potential for autogenesis can be maintained even in non-permissive conditions. Although such a molecular configuration is only an autogen with respect to a specifically supportive environment, it can nevertheless persist structurally across diverse and unsupportive environments, to later be reproduced in a supportive one. In this way, autogenic ident.i.ty maintenance transcends any specific context dependence. In its inert closed state, an autogen can maintain this potential across vast epochs of time and through diversely non-supportive contexts. This then provides a degree of autonomy from context that is again a distinctive quality of living but not non-living dynamical systems.

An autogen's individuality is strangely diaphanous in one interesting sense. When closed and complete, an autogen is inert and yet when broken open and actively forming new const.i.tuents, it is merely a collection of molecules dispersed into a larger molecular milieu. In other words, it is a bounded individual only when inert, and actively self-generating only when it is no longer a discretely bounded material unit. This further demonstrates that what const.i.tutes an autogenic "self" cannot then be identified with any particular substrate, bounded structure, or energetic process. Indeed, in an important sense, the self that is created by the teleodynamics of autogens is only a virtual self, and yet is at the same time the locus of real physical and chemical influences. This was also a feature recognized by Francisco Varela in his conception of an autopoietic system, even though he did not recognize how synergistic reciprocity of self-organizing processes could produce this. He describes this virtual self as follows: What is particularly important is that we can admit that (i) a system can have separate local components [for] which (ii) there is no center or localized self, and yet the whole behaves as a unit and for the observer it is as if there was a coordinating agent "virtually" present at the center.8 The generic, autonomous, and diaphanous character of autogenic systems makes this a functional property, not a material, chemical, or energetic property. An autogen is a precisely identifiable source of causal influence because it generates and preserved dynamical constraints-the basis for thermodynamic work. But, as we've seen, constraint is the exemplification of something that does not occur. So, in this sense, an autogenic system confronts us finally with an unambiguous absential quality. This is the essence of teleodynamical causality. By examining this model system more closely, we will be able to demonstrate how ententional processes acquire this seemingly paradoxical character of efficacious absence.

AUTOGENIC EVOLUTION.

Because autogens are capable of self-replication, they are also potential progenitors of autogen lineages. An autogen lineage will increase in numbers so long as there are sufficient substrate molecules in the surrounding environment, along with sufficient molecular agitation to periodically disrupt autogen integrity, but not so much agitation as to disperse their contents more rapidly than they can rea.s.semble. This is the first condition for natural selection.

The second condition is compet.i.tion among these lineages. This will occur spontaneously as well because the multiplication of each lineage is dependent on the same (or catalytically similar) molecules in the surrounding environment. Different lineages are therefore in compet.i.tion for these substrates, as well as for persistence against the relative disruptive influences potentially present in the environment-and to an extent necessary for lineage growth.

The third condition is variation among these lineages. Let's begin with the simplest case: a single type of autogen. Although autogens tend to self-reconst.i.tute and thus reestablish the molecular structures predisposed by the catalysts and sh.e.l.l molecules that produced their progenitor, because this must occur via partial breakup and reclosure of the sh.e.l.l, it is likely that each new sh.e.l.l will additionally incorporate other molecules present in the local environment, some of which may continue to be pa.s.sed on via future divisions. For the most part, these will tend to be innocuous inclusions that might only decrease the relative concentrations of active catalysts. Higher or lower concentrations of these relatively neutral molecules will tend to produce slight differences in rates of reconst.i.tution or sensitivity to dissociation between lineages. If, however, incidentally included molecules are uncorrelated with any allosteric specificity of autogen functions, their addition to or loss from a lineage will be a matter of chance. Some fraction of incidentally incorporated molecules will have allosteric similarities with functional catalysts, their products, or their substrates. These will have a tendency to interact directly with them and will affect functional attributes of autogen chemistry. Their possible effects include directly interfering with or augmenting the rate of catalysis, affecting container formation, altering container stability, influencing molecular diffusion during disruption, providing linked parallel catalytic steps in the cycle that enables utilization of alternate or slightly variant molecular substrates, and so on.

Incorporation of functionally interactive molecules will make a given lineage relatively more or less successful in propagation. Autogen lineages containing different types of molecules that augment function in any of these ways will tend to out-reproduce others. This will only have a permanent selective effect on a lineage when another condition is met. Persistent inclusion of these divergent molecular types within a lineage will depend on being synthesized as part of an extended or parallel autocatalytic cycle. By these means, increased complexity and reliability of autocatalysis, improved containment, and utilization of special structural and energetic features of other molecular types (e.g., metals) could evolve by differential lineage propagation.

Together, these considerations show that although autogens are incredibly simple molecular systems, their self-reconst.i.tution properties in favorable environments spontaneously bring into being the systemic conditions that are sufficient to initiate a persistent, if weak, form of natural selection. So, in an environment where autogen reproduction is likely, autogen evolution is also likely. In molecularly complex environments, autogen lineage compet.i.tion for resources will tend to lead to the evolution of variant lineages differentially "fitted" to their local environments. This satisfies all necessary and sufficient material and logical conditions for natural selection, despite occurring in a system lacking many core attributes of life, including genetic inheritance in the biological sense.

Having established the logic of autogen evolvability, we can briefly consider certain special cases and features of this process that diverge from the norm of natural selection among organisms.

One potential source of lineage difference is independent emergence. It is possible that entirely independent and distinct types of autogens could arise in slightly differing environments, each giving rise to lineages that eventually overlap in their environments. Such independently arising lineages need not have overlapping substrate needs, and so would not necessarily be in compet.i.tion. Multiple parallel threads of autogen evolution are possible, and perhaps likely, since a molecular context that is sufficiently rich to give rise to one type probably has the potential to give rise to other variants as well. This would not necessarily produce a natural selection dynamic, except to the extent that there was chance overlap of some substrate. Even so, their independent effects on other aspects of the environment might eventually lead to indirect effects that influence relative propagation rates in some more generic way (e.g., by effects on molecular diffusion).

This is crudely a.n.a.logous to the way different species in an ecosystem affect each other's evolution, even if not directly interacting. But the pa.s.sive and partially destructive nature of autogen propagation, which is quite unlike most living cellular reproduction, also introduces other potential avenues of evolutionary interaction. The most direct form of interaction could arise by mixing, as a result of inclusion of the components of one type of autogen within another. Since containment and co-localization of essential catalysts are somewhat generic features of autogen dynamics, there are many opportunities for linkages to develop between lineages with very different origins, and even for inclusion of the entire catalyst set of one autogen type within another. This is like endosymbiosis, now recognized as an important, if perhaps rare, source of novel synergistic functions in evolution, or lateral gene exchange, which is common in bacterial and viral evolution. The incorporation of complementary self-reconst.i.tuting systems within other self-reconst.i.tuting systems with somewhat different substrate requirements, catalytic dynamics, and structural self-a.s.sembly conditions could be a significant source of complexification and adaptation. The somewhat unregulated cycles of breakup and re-enclosure that autogen propagation depends on are in this way more a.n.a.logous to virus reproduction than to cellular reproduction, and so may share other characteristics with virus evolution as well.

Natural selection is the aggregate dynamic that arises out of the interactions of large numbers of unit systems with these properties. Different lineages of autogens each maintain a thread of mnemonic continuity and ident.i.ty, even when mixture occurs. Irrespective of such lineage convergence effects, when they are separated in different lineages, variant autogenic mechanisms will be in compet.i.tion with respect to resources available in surrounding conditions. This makes each individual autogenic mechanism also a representative of a specific correlation between its intrinsic dynamic topology and features of the larger environment.

So, with autogens, there are multiple interlocking units of evolutionary individuation: the lineage, the autogen as autonomous system, and the reciprocally self-maintaining dynamical synergy, though the latter will likely grade into an autogenic unity once independent reciprocal dynamics intertwine into a single integrated system. Living organisms derive their peculiar causal dynamics-their seeming end-directedness, functional logic, superficial reversal of the second law of thermodynamics, and adaptability-from this upward shift in what const.i.tutes a unit of reproductive and evolutionary individuation.

The foregoing a.n.a.lysis has shown that natural selection emerges from the dynamics of reciprocally reinforcing self-organizing processes. This self-reconst.i.tutive dynamical synergy is the essential ingredient that precedes and underlies life. This in turn suggests that natural selection is ultimately an operation that differentially preserves certain alternative forms of morphodynamic processes compared to others, with respect to their synergy with one another, and with respect to the boundary conditions that enable them. Selection is not then fully defined only with respect to replication of genetic information. As Kant recognized, a self-maintaining "formative power" is critical. And this requires processes that generate, preserve, and propagate constraints. Morphodynamic processes are the only spontaneous processes that generate and propagate constraints, and autogens demonstrate that reciprocity between morphodynamic processes can preserve and replicate constraints.

THE RATCHET OF LIFE.

Autogens may be subject to natural selection and evolution but they are not alive in most senses. They lack the majority of attributes a.s.sociated with living organisms today. Most significantly, they are effectively pa.s.sive-though not inert-structures, because they do not actively acc.u.mulate and mobilize energy within themselves that can be used for self-repair and reproduction. The energy driving their self-perpetuating dynamics may derive from an environmental source (such as the heat of volcanic vents), or it may be obtained by breaking the molecular bonds of large substrate molecules. In any case, autogens are parasitic on their environment, even for the initiation of replication, in much the same ways that viruses are. Unlike viruses, however, autogens reproduce without the help of other organisms.

In this respect, autogens are not quite "autonomous agents," in Stuart Kauffman's sense. For Kauffman, the most basic characteristic of living systems is their ability to "act on their own behalf." Although an autogen does not "act" in a self-animated sense, it nevertheless fulfills the two criteria that Kauffman uses to define what he means by autonomous agency-criteria that he argues are the basis for this self-directed activity. It ha