Evolving Thoughts

What is “life”, at last

Recently, that is since 1975 or so, the view has arisen that a living thing is something that satisfies several conditions.

In 1966 George C. Williams introduced the notion of an “evolutionary gene” in his Adaptation and Natural Selection, which was, he said, a “cybernetic abstraction”. This idea was taken up by Richard Dawkins in his The Selfish Gene.

Dawkins posited that evolution had some necessary and sufficient criteria:

There had to be replicators with the following properties:

  • Longevity (over evolutionary time)
  • Fecundity (more made than can survive)
  • Fidelity (nearly perfect copying)

and these had to be contained in vehicles, which were the bodies that they “made”. David Hull, a philosopher, revised this to interactors, which were ecologically active agents, that determined the fitness of the replicators over evolutionary time.

The Hull-Dawkins Distinction, as it is now known, is considered by many to be the foundation for a “Universal Darwinism“, a generalised account that applies to any evolutionary process. Dawkins himself applied it to cultural evolution, and Hull has published with an immunologist and an ethologist a general theory of selection (Hull et al 2001).

If things evolve, then, they have these properties, and if they have these properties, then they are Darwinian things, and hence are prima facie alive. On a Universal Darwinian view, if it evolves, then it has the properties that are sufficient for life. So cultural ideas are like “mind viruses” that “get” themselves replicated by parasitising our brains.

Problems with these criteria are obvious as a definition of “life”: ideas are obviously not alive. They do not have the properties of living systems like metabolism. Nor do Animats and genetic algorithms.

So we are left at the end with a series of problems.

  1. If we include overly general criteria, we include nonliving things
  2. If we restrict our criteria the way Orgel does, we exclude non-terrestrial life
  3. If we try to generalise the physical properties of terrestrial life in a way that will include non-terrestrial biology, we run the risk of falling into one or two of the previous problems.

So, how to proceed? Can we define “life”? It depends a lot on how we think general terms get their meaning. Traditional philosophy thought, following Aristotle’s example, that if we only defined them clearly and unambiguously enough, we could include what we wanted to include and exclude what we wanted to exclude. This may be the problem.

Language in science gains its meanings in all kinds of ways. One is for a term to derive its meaning from a theoretical context. Hence “mass” is defined in one way in Newtonian contexts, and a different way in Einsteinian. But biological terms are usually applied first to exemplars, prototypical examples that we are familiar with, and then applied increasingly generally until they break down.

This explains why so many biological terms, that are not theoretical, break down at the periphery. We devise terms for things that are unproblematic and then expand the domain over which they apply until sufficient difficulty arises. One of the obvious features of our experiential world is that things live. So we take exemplary cases and try to expand the use of the term until we get to viruses, prions, and eventually extraterrestrial biology. The problem is that we are using analogy, not homology.

All life on earth, barring Carol Cleland’s possible but as yet undetected parallel biologies, evolved from a single origin of life event, and subsequent elaboration and reticulation of the lineages that evolved from it. So to define “Earth Life”, just point at the two most diverse examples one can find, and say “life is anything that evolved from the last common ancestors of those things”. This is definition by evolutionary homology, ostensively anchored in the world by pointing.

But analogous definition is “something that is like that” where “that” is some clear case that you point at. This is what motivated Aristotle. The trouble is that, as we have seen, “that-ness” is something too vague or inclusive to clearly identify what it is we are talking about (unless we restrict the “thats” to an evolutionary group such as metazoans, or fungi, or plants, etc.).

Analogy is not a good way to proceed in scientific definition, because it is subjective. It depends not on the objective properties of the things defined, but on the criteria that the researcher thinks is significant. In short, it defines facts about scientists, not the subjects of study. The use of theoretical definitions, however, attempts to avoid this because theories are supposed to be rigorous descriptions of objective reality. Hence definitions of “electron” work, because the theory has been refined to the point where it is entirely adequate for the domain of physics.

In biology, however, most definitions must necessarily be ostensive and phenomenally based, unless we apply a phylogenetic definition. This means, though, that an ostensive definition based on common ancestry excludes life on other planets. And now it is time, perhaps, to qualify what it is that is being defined: the last common ancestor of our two most unrelated organisms defines, extensively, only “Life on Earth”. It doesn’t define “life” universally. If we find that Jupiter’s clouds do enough of the things that we ordinarily count as living by analogy (reproduction, ecological interaction), then we will apply our pre-existing word “life” to them, and thereafter have to use an extensional definition for “Jovian Life”, and so on.

Will there be a general theory of life? I doubt it. Or rather, I think we already have that general theory – it’s the laws of chemistry, thermodynamics, and physics in general. But knowing all about physics won’t tell us what sorts of particular tricks the physical world has for doing complex stuff like reproducing. To know that, we have to encounter it first, and then work out how it happened according to the rules of What Chemistry Can Do. And Physics. And Systems. And so on…

Hull, D. L., R. E. Langman, and S. S. Glenn. 2001. A general account of selection: biology, immunology, and behavior. Behav Brain Sci 24 (3):511-28; discussion 528-73.

First post in this series

Second post in this series


  1. #1 Benjamin Franz
    September 13, 2007

    Problems with these criteria are obvious as a definition of “life”: ideas are obviously not alive. They do not have the properties of living systems like metabolism. Nor do Animats and genetic algorithms.

    A number of researchers in the field of artificial life would differ with you as to the ‘obviousness’ of your assertions.

    To start with, ‘Ideas’ and computer life do ‘metabolise’. That their physical base is not always ‘wet’ doesn’t mean they are not converting available physical energy into useful (to the organism) work through physical actions. Their physical basis is subtle, but not non-existent. A ‘bit’ of computer memory has an actual physical base.

    Interestingly, DNA, when looked at from a purely information viewpoint, is just a form of digitial data storage. Precisely two bits per codon with an error correcting system on the third codon that ensures that an error in the third codon has a 2/3 chance of causing no transcripton errors. Digital information is the genetic basis of all known forms of life.

    ‘Wet’ life is nothing more than a sub-type of self-replicating machines.

    BTW, it’s not nice that adding any hyperlinks to references causes SB to refuse comments to your blog.

  2. #2 John Wilkins
    September 13, 2007

    Hyperlinks should work. I sometimes get an email asking for approval for the response, but they ought to get through.

    I once felt that ALife and memes and the like were alive in some relevant sense. But now I think they are just things that evolve, and there’s no need to extend the term “life” to them. Evolvability is a necessary but insufficient condition, in my view. But if you accept that definition, it’s at least consistent – I just think it is not a good use of the terms.

  3. #3 Heleen
    September 13, 2007

    Tibor Ganti and Eors Szathmary are sadly missing in this discussion. When Zimmer writes Bedau has been pondering just what those milestones might be. To qualify as fully alive, he argues, a system needs three basic features. Life needs a container; it needs a way to encode and replicate information; and it needs a way to capture and use energy., Zimmer might be quoting an opinion as favoured by Bedau that is already generally known, quite old and accepted, rather than a novel idea by Bedau. The clue should be in the 1999 book by John Maynard Smith and Eors Szathmary, The origins of life : from the birth of life to the origin of language, as well as in their larger 1995 book. These ideas go back to Tibor Ganti.

  4. #4 RBH
    September 13, 2007

    Your rejection of the Hull-Dawkins Distinction depends on introducing additional criteria. Accepting the criteria of the Distinction — longevity, fecundity, and fidelity of inheritance in interactors — then ideas, animats, and the critters in evolutionary algorithms are “alive”. Only by introducing additional criteria to justify your “obviously not alive” judgment can you reject the Distinction. And you do introduce additional criteria: you say “they do not have properties of living systems like metabolism.” Well, some living systems have metabolisms and some critters in evolutionary algorithms have metabolisms (they consume computer cycles), but that’s irrelevant unless one adds “has metabolism” or “processes inputs to produce outputs” or some such locution to the criteria.

    I’m perfectly happy with a fuzzy boundary such that the artificial agents that evolve in my genetic algorithms, critters that are long-lived in evolutionary time, are fecund, that compete for living space, and that process inputs to produce outputs influence their probability of successful reproduction (and whose success in their environment determines my economic well-being) are “alive” in the most general (and important) senses of the term.

    Is there fuzz at the boundaries? Sure there is, but there is no requirement that the world be Aristotelian. I’m comfortable with the notion that there are ‘semi-living’ things, too.


  5. #5 Larry Moran
    September 13, 2007

    If things evolve, then, they have these properties …

    No they don’t. Things can evolve without making more than survive and things can evolve with sloppy copying.

    I think you (or Hull & Dawkins) confused evolution and natural selection. I admit that things can’t survive forever if they don’t make more copies of themselves than survive but at least in the short term they can evolve.

    As you might guess, I’m uncomfortable with a definition of life that relies on an adaptationist worldview. I think I can imagine life that changes by inheritance of acquired characteristics. On that planet Lamarck is God.

  6. #6 Boronx
    September 13, 2007

    Memes don’t fit one item of the definition:

    “Fidelity (nearly perfect copying)”.

  7. #7 DrA
    September 13, 2007

    Why focus on the replicator? Isn’t life a process that just delays increasing entropy? Does anything other than living organisms capture energy and lock it up for a period of time in molecular bonds before it ultimately continues dispersing as heat?

  8. #8 John Wilkins
    September 13, 2007

    Please understand that I am reporting here, not promoting the HDD account as a definition of either life or evolution.

    I think that all that is needed for evolution is reproduction, and reproducers can be a fair bit less exact than replicators (in fact the latter are a subset of the former category). If you have replicators, then you get evolution, but the inverse inference is not true.

    Ganti and Szathmary (I can’t put the accents in with ease) are missing, and Heleen identifies the reason – this is assumed by Bedau. I personally like the chemotron as a generalised account (but it is included under the chemical definition and the metabolic definition), and the compartmentalisation and error catastrophe account would take too long to explain (oh, OK, too long to look up and assimilate for a blog post).

    And memes are not (necessarily) replicators, yes. But if they are reproducers, one can get past Atran’s and Sperber’s criticisms.

  9. #9 Caledonian
    September 14, 2007

    Is there fuzz at the boundaries? Sure there is, but there is no requirement that the world be Aristotelian.

    Except, of course, for the requirement that the world be describable in terms of things that are true and false.

    So if there’s a “fuzzy” boundary, where things cannot be classified even in principle, then there’s actually no boundary at all and the concepts you’re dealing with are incoherent.

    As for the definition: there seems to be a confusion between establishing a definition, and attempting to generate a definition that describes a pre-existing category. If you have sets of things that you recognize as ‘alive’, and you want to simplify their definition, the very first thing you’d want to do is specify what’s in your category, so that you can describe it properly.

    The fact that these steps haven’t been taken is telling.

New comments have been temporarily disabled. Please check back soon.