Over at Discover, Sean Carroll has a nice post on thermodynamics, free energy and the origins of life. It’s a good intro, but in the course of it he remarks:
Obviously there is a lot missing to this story, and much of it is an absence of complete understanding on my part, although some of it is that we simply don’t know everything about life as yet. For one thing, even if you are a metabolism-first sympathizer, at some point you have to explain the origin of replication and information processing, which plays a crucial role how we think about life.
Well, now, that is a red rag to an information-eliminativist, isn’t it? How could replicators and information processing evolve from a “metabolism first” account of the origins of life? The answer to that is exactly what Carroll is extolling: thermodynamics.
A while back, Eric Schneider and the late James Kay published a paper “Life as a manifestation of the second law of thermodynamics“. They point out that if living systems are seen as autocatalytic cycles, what Manfred Eigen and his collaborators called a hypercycle, life can originate as self-organising hypercycles just from the fact that where you have a heat or energy gradient, you will have a region where the flow of free energy can spontaneously cause organisation. Eigen’s hypercycle has no replicators. A replicator, on the view devised by Richard Dawkins, is any part or section of a process that meets three criteria:
… longevity, fecundity and copying-fidelity. Longevity means longevity in the form of copies through descent (Dawkins 1982a:84, Hull 1980), although the stability of tokens is included in the definition in The Selfish Gene (Dawkins 1976:18, 1999:17).
A gene is a replicator, but Dawkins permits there to be other kinds (including memes), but with respect to the origins of life, there are only two alternatives: replicators came first, or they evolved from some other process. However, here’s my problem with Dawkins’ standard view:
We have no evidence about what the first step in making life was, but we do know the kind of step it must have been. It must have been whatever it took to get natural selection started. Before that first step, the sorts of improvement that only natural selection can achieve were impossible. And that means the key step was the arising, by some process as yet unknown, of a self-replicating entity. Self-replication spawns a population of entities, which compete with each other to be replicated. Since no copying process is perfect, the population will inevitably come to contain variety, and if variants exist in a population of replicators those that have what it takes to succeed will come to predominate. This is natural selection, and it could not start until the first self-replicating entity came into existence.
My problem is this: to posit that some molecule just acquired the capacity to replicate is to posit a scientific miracle. It’s a bit like suggesting that a molecule might just acquire the ability to act as a transistor. I do not like scientific miracles – they strike me as an admission of failure. So I want to see that we could explain how replicators themselves evolve. Of course, if replicators are the sine qua non of evolution, that without which evolution is impossible, then we need some other improbability reducer, such as the clay substrate hypothesis of Cairns-Smith Dawkins has several times discussed. There are others. But they all seem to me to end up explaining at best only one aspect of replication, and most certainly not how a replicator might become embedded in a cycle of metabolic and reproductive processes.
This sine qua non exclusivity for replicators is, I think, mistaken, and many years ago I wrote a paper that never got sent anywhere in which I argued this, with two collaborators, Ian Musgrave and Clem Stanyon. We always intended to do a simulation to prove the concept I am about to outline, but never did, and more detailed accounts have since been published. In particular, a challenge to the necessity for a replicator before you get natural selection has been raised by Jim Griesemer with his notion of a “reproducer“. What I want to do here is show how selection upon non-replicative hypercycles as reproducers can give rise to replicators.
Griesemer introduces reproducers in contrast to Dawkins’ and Hull’s refinement of, the replicator:
In contrast to the Dawkins-Hull account of replication, my view is that whether or not “the” parental nucleotide sequence is instantiated in offspring is irrelevant to an analysis of reproduction. Hence, I will claim, it is irrelevant to the analysis of evolution as a process, however much this instantiation in offspring bears on evolutionary outcomes by determining the degree to which descent with modification causes resemblance among members of a lineage. In a word: one cannot produce process out of function alone. If the mechanism of DNA replication were indeed essential to reproduction in general, then Dawkins’s copying process might support a suitably general analysis of evolutionary benefit. But since this mechanism is contingent, the criteria Dawkins attributes to replicators such that they are the sole beneficiaries of evolution fail. It is not necessary that reproducers be replicators in Dawkins’s sense. Reproduction could occur by some other mechanism and the reproductive entities would nevertheless be beneficiaries in the relevant evolutionary sense. [95f]
Indeed. If reproduction is the basis for life, then replication becomes one of a number of ways in which things can reproduce. In a paper published earlier than the one cited above, but which was written later, Griesemer clarifies his view:
Thus, rather than thinking of ‘replicator’ as a generalization of the gene or the genotype concept, replicators – units of replication – are a special class of inheritors – units of inheritance – which in turn are a special class of reproducers – the units of reproduction – which in turn are a special class of multipliers. There is a hierarchy of concepts of which ‘replicator’ is the most specialized…
On Griesemer’s account, a reproducer is a physical object that “materially overlaps” its progeny. A replicator is a special case of physical object that has evolved a coding system and developmental process, and which causes, physically, the progeny to generate. By contrast, Dawkins’ and the general version he is criticising treats a replicator as an abstract entity, a form of information, which is related not by causal lineages but by similarity relations. I think this is a little unfair to the Dawkins view; he clearly sees replication as a form of causal process, but the problem is, what kind of causal process?
In the philosophy of mind and language, it is often – I think rightly – asserted that abstract sets do not cause anything, but only token of sets. Another way to express this is that abstractions don’t cause things, but actual objects that are instances of the abstraction. For example, the set of bowling balls knocks over no pins, not even abstract pins. Actual bowling balls do. So if replicators are abstractions, they can do nothing. To say otherwise is to adopt the metaphysics of Aristotle, or Aristotle as interpreted by the medieval scholastics, in which form (the Greek for which is eidos, cognate with ideai) accounts for the properties of things, not what they are comprised (substance) of. Modern physical ontologies reject this hylomorphism in favour of what we now call physicalism, and to assert that information is itself a physical property is, I believe, deeply mistaken, and a kind of hylomorphism.
So if Griesemer is correct about this, and I strongly think he is, subject to a qualification I will explain in a minute, then replicators must be physical objects, and the Dawkins-Hull view of replication, so far as it is based on information being a causal property (which I think one can reinterpret otherwise), is wrong. I need to do some extra work to decide whether that notion of replication in fact needs to treat information as a property. Hull, at any rate, seems to have changed his approach at various times, and is more ontologically neutral than Dawkins.
But how do we get causal replicators? In my next post on this topic, I will argue that it happens through a selection-like process on reproducing hypercycles, in simple thermodynamic terms.
Now to that qualification. In his recent book, Darwinian populations, Peter Godfrey-Smith has argued (p83f) that the “material overlap” requirement is not needed as it is too narrow a condition (Jim thinks that reproducers must include parts of themselves in their progeny). I agree; a reproducer must in fact merely impart physical quantities rather parts. Suppose the clay hypothesis were correct – what is imparted here is the configuration of the molecules, in that the prior crystalisation acts as a template for the subsequent layer. What is imparted is not “information“, whatever that might be, but structure through the straightforward physical interactions of molecules via weak and strong interactions in a thermal medium; in other words, thermodynamic properties. All that is needed is a surface.
So, when I get time, I’ll outline my and my collaborators’ argument for how replicators in this sense might evolve through selection.
Dawkins, Richard. 1976. The selfish gene. New York: Oxford University Press.
———. 1982. Replicators and vehicles. In Current problems in sociobiology. Cambridge UK: Cambridge University Press:45-64.
Godfrey-Smith, Peter. 2009. Darwinian populations and natural selection. Oxford: Oxford University Press.
Hull, David L. 1980. Individuality and selection. Annual Review of Ecology and Systematics 11:311-332.
Schneider, Eric D. and James J. Kay, 1994. Life as a manifestation of the second law of thermodynamics. Mathematical and Computer Modelling 19(6-8): 25-48