Evolving Thoughts

The meaning of “life”

Carl Zimmer has one of his usually clear and precise articles on recent work on the nature of life, focussing on the work of Carol Cleland, who is at the National Astrobiology Institute, despite reduced funding for actual science by the present administration. I met Carol last year when we both spoke at the Egenis conference on the philosophy of microbiology.

Carol argues that we do not have a theory of what life is. She may be right. One of her arguments is that if there were multiple “origins of life” events that used different chemistry, we may not even be able to “see” the others because we’d be using the wrong chemical assays, and she (and others) point to the possibility that the Viking Lander in 1976 actually did find Arean life, only it killed it because we were testing for water-based life, and it may instead be based on hydrogen-peroxide, which is liquid at the temperatures and pressures on Mars.

I want to do a slightly different thing here. Instead of asking what life is, I want to ask what “life” means. That is, what do the word and associated concepts have as content? What do we mean when we use the word “life”?

As always in these things, it goes back to Aristotle. In De Anima (peri psuche, or On the Soul) Aristotle lists an ascending series of “souls” or “motive forces” that characterise living things. It’s very interesting that he doesn’t make a sharp distinction between living and nonliving things, as indeed nobody did until the modern discipline of biology got started. Linnaeus, for example, applied the very same classification principles to living species as to nonliving (mineralogical) species.

The “souls” that Aristotle assigned were, in order of decreasing generality:

  • The “nutritive soul”, or the ability to nourish and grow
  • The “sensitive soul”, or the ability to react to one’s surroundings
  • The “appetitive soul”, or the ability to desire something
  • The “locomotory” soul, or the ability to move independently
  • The “rational soul”, or the ability to make inferences and reason

These became, in the scientific discourse, metabolism (literally, “to cast beyond”), irritability, goal-directed behaviour, motion, and nervous systems, but overall the criteria for life haven’t radically shifted since Aristotle. This view is sometimes called psychism, from the Greek word psuche which gets translated as “soul”, but it is better to think of psuche as a “motive force” than something like the theistic notion of a soul.

These definitive criteria for “life” are, of course, based on observation of prototypical living things. If you are in the macroworld, as we are, then cattle, horses, birds, and humans are the obvious exemplars of living things. Extending the properties shared in common to plants, something like Aristotle’s criteria are sure to be formulated, because it’s clear that plants don’t (often) move, and so on, but every living thing eats and grows. As a side note, it’s interesting that Aristotle doesn’t include reproduction, although he discusses “generation” quite a lot. Perhaps he thought this was just something that all things, living or not, have, a kind of causal power.

Aristotle’s, and modern science’s (such as the Astrobiology program at NASA) approach has been to take an a priori approach, classifying the properties of life “top-down”, as it were, from large scale objects to other forms of life that don’t, quite, match these criteria. The debate over whether viruses are living is an old one, for example. They lack metabolism and all of the other criteria, but the do reproduce, and in the modern scheme of biology, reproduction is considered crucial for life. Reproduction is perhaps the only “modern” criterion that has been added to Aristotle’s scheme.

This top-down approach means that we begin to form our concepts based on our naive experience. Buffon, the famous 18th century natural historian, began his classifications from horses, which were crucial to the daily life of a French nobleman (which he was, and many of his audience were also). But biology is supposed now to be independent of the interests of French noblemen, and indeed human social values. It seems that the very foundation of biology is still enmired in social values and the cognitive limitations of humans, though. We start with living things very like ourselves, and generalise from that by subtraction. Sometimes we even add organisms to those categories – much work on the cognitive symbolic reasoning of chimps, birds and cetaceans includes them under Aristotle’s “rational soul”. But our categories and criteria themselves have not changed apart from the addition of reproduction.

Why reproduction? In part this is because the past two centuries have seen enormous progress on that process, whereas previously it was opaque to research. In part it is because reproduction is the crucial process in evolution according to the Darwinian scheme. In part it is because, I believe, we are enamoured of information transmission, and have been since the telegraph and semaphore was invented, and the replication of genes is seen widely (but erroneously!*) as the exemplar of that in biology (and hence the debate whether viruses are alive or not).

But suppose that life really is just another facet of What Chemicals Can Do. Suppose there is no “theory of life”, just a general theory of physics and chemistry – in particular the laws of thermodynamics and molecular recombination – and a historical or contingent account of what some chemicals have done on one planet in the universe. This would both bolster Carol Cleland’s claim that there may be other “biological” forms of “life”, and undercut her desire to see a universal definition of that term.

Way back in the 19th century, W. Stanley Jevons, in his Principles of Science noted that we begin classifying on phenomenal grounds. First we notice all the large animals, then we notice the mammals, the birds, the fishes, the worms, the arthropods, and so on, and we continue until we reach the limits of discrimination, and we have species. However, working up from the bottom, the so-called “reductionist paradigm”, doesn’t seem to give us the organisms, let alone the species or higher groups. Physics is physics and chemistry is chemistry. There is no more a theory of life than there is a theory of high school chemical experiments. Hmmm…

Learning about organisms directly is what experimental and field biology is about. Molecular approaches deal with a range of phenomena, such as metabolic processes or immunological cascades, at that level. You couldn’t predict that a particular kinase is the trigger for an immune reaction from a knowledge of the molecules involved,because what role they play in individual organisms depends on everything else in those organisms. Perhaps we will formulate a general theory of systems that covers living things, but it will be a theory of systems not of life. There are going to be things that are systems but not alive. Our definition of the word “life” is based on what we see, on things that are exemplars of life. There may be no theory here, just observation and history.

* So I reckon, anyway.


  1. #1 coturnix
    September 6, 2007

    “…but erroneously!*….”
    is asking for a footnote….

    How do ‘energetics’ definitions of Life fit into your essay? If something, at least at some times, takes in, stores, uses, and dissipates energy and does it against the energy flow imposed by the environment, than it is pretty much alive. Or is it?

  2. #2 coturnix
    September 6, 2007

    Oh, I was too fast – you just added the footnote.

  3. #3 John Wilkins
    September 6, 2007

    Yeah, I realised as I hit “Save”…

  4. #4 John Wilkins
    September 6, 2007

    Oh, and I’ll continue this theme in a subsequent post, so don’t despair about energetics.

  5. #5 Caledonian
    September 6, 2007

    What do we mean when we use the word “life”?

    Why are you willing to claim that there’s no definition of ‘science’, but are trying to investigate the definition of ‘life’?

  6. #6 R. Landau
    September 6, 2007

    “the replication of genes is seen widely (but erroneously!*) as the exemplar of that in biology”

    I’m curious as to why you think this is erroneous. Do you mean that replication of genes is not transmission of information? Or that genes do not contain information in the pertinent sense? In the later case, it seems to me that a genome if chock full of information.

    On an obvious level, the genome has the information necessary to construct a particular life form, with all the various pieces and parts that go into it. But on a perhaps slightly less obvious level, the genome contains information about the type of environment the creature comes from.

    For example, if we could read a strand of DNA and determine that it codes for an organ that is designed to filter oxygen out of a gas, we can make an educated assumption that the creature comes from an environment where there are gasses with oxygen in them.

    If we could read another strand and see that it coded for an organ sensitive to certain spectra of light, we can assume that the creature comes from an environment where these spectra occur.

    So, on one level, the DNA tells how to make a creature, but on another, the construction of the creature says something about the environments it is expected to be in. So the DNA contains information on the environment that gave rise to itself.

    The former case might be a little more tricky, but it still seems to me that information is being passed on.

  7. #7 John Wilkins
    September 6, 2007

    The reason I think genes aren’t information is that you simply cannot read them (except for primary sequences of proteins, which is better described as catalysis). They do not behave anything like any sense of “information” of which I am aware. DNA doesn’t tell anything how to do anything; it is used as a resource to construct the building blocks of an organism, not the organism itself. It acts, as it were, as a rate governor for that purpose. If I give you an arbitrary DNA sequence, even if it is several reading frames long, you can’t say one damned thing about what it “codes for” above the protein level (and often, not even that) unless you already know a whole lot about the kind of organisms it comes from and how they typically develop.

    DNA also doesn’t “code for” information about the environment as such. It regulates some changes in the overall development of organisms (again, if you already know the rest of the organism) that are usually effective in a given environment, but transplant it to another organism and all bets are off if it isn’t a conserved homologous gene.

  8. #8 Jim Thomerson
    September 6, 2007

    I don’t think there is life, the noun. Rather there are only living things. To use life as a noun is, I think, vitalistic regardless of good intentions. I have described life as an emergent property of a properly organized complex system. Not sure there is whole lot of information content in that statement except to characterize life as a property arising from organization.

  9. #9 MartinDH
    September 6, 2007

    Isn’t this just another example of the search for a bright, dividing line? This time between life and chemistry.

    My feeling for this is that life must have a hereditable, reproductive capability.

  10. #10 Size
    September 6, 2007

    Nicely written. Perhaps part of the reason we’ve had a hard time defining so many other relevant terms (“individual” and “species” are two recent topics that come to mind) is because we’re lacking a theory that adequately handles systems, including life. Perhaps “life” is merely a convenient word for dealing with certain kinds of systems.

    My suspicion is that life is an emergent property rather than a state of being that can be defined as being simply “on” or “off”.

    On the other hand, I sort of naturally tend to toward the perspective that “we’ll know it when we see it.” Obviously, as in the case of viruses and prions, not even that is a given.

  11. #11 Caledonian
    September 6, 2007

    The reason I think genes aren’t information is that you simply cannot read them (except for primary sequences of proteins, which is better described as catalysis). They do not behave anything like any sense of “information” of which I am aware.

    Then you desperately need to stretch your horizons, because information is precisely what DNA transfers.

    It’s trivially easy to write a computer program that you “simply cannot read”, in that merely reading the code straight through doesn’t provide a picture of how the processes it describes will unfold, but I hope you wouldn’t dare claim a computer program isn’t information.

    If you really want to delve into this issue, I suggest developing a working layperson’s understanding of the mathematics of computation and information theory.

    As for ‘life’, the word is used to describe many different concepts, most of which have no specific operational relationship to the others. Probably the best functionial definition of the term is “a collection of processes that actively perpetuate themselves”. Tornados and forest fires are perpetuating processes, but they don’t actively create conditions favorable to their perpetuation by seeking out resources. Bacteria search our the resources they need, so bacteria are living. So are viruses.

  12. #12 John Wilkins
    September 7, 2007

    Cal: I’m not going to engage you in a slugging match over this. Suffice it to say that I do have a fair amateur knowledge of information theories, and I still have an argument, one that does not rely on the “it’s just obvious” premise, that genes do not have information. I have blogged about this in the past:

    Instruction and Information

  13. #13 Torbj�rn Larsson, OM
    September 7, 2007

    Zimmer’s article has triggered several nice posts like this one. [I see John has already seen the meat of what follows. But at least I happened to throw John a bone. :-P]

    It represents only a single data point, … [Cleland about life on Earth, in Zimmer’s article on Seed.]

    I don’t think this is necessarily true for the case and characteristics considered. Viruses and cells from different extant and extinct domains could have crossed the Darwinian threshold several times from a progenotic state. Also, early total extinctions are consistent with the observation of early and so likely easy abiogenesis.

    I get the impression that these crossings are initiated by diverse selfish elements, a mechanism that is often observed in genes.

    My layman conclusion from this is that evolution, especially past a progenotic state, is a competitive [sic!] and robust process which we should expect to be common elsewhere.

    Maybe we will meet singular existences, biological or mechanical, that aren’t described thusly. But I believe the way to bet is that they are rare and bound on an extinction path.

    I generally champion definitions that are inspired by theory since they correlate better with it. Therefore I currently try on for size the following conception of life (to use John Wilkins terminology on species):

    An organism is the unit element of a continuous lineage with an individual evolutionary history.

    The model for the definition is an organism as the current slice in a continous process. Thus combining the idea of life as individual and life as process with evolution and an implicit assumption of a hereditary mechanism, and a robust definition of organism. Quite a few birds knocked down with one stone.

    As I understand it this definition excludes organelles and such replicators as prions because they have entered dependent niches, as they are subsumed into an organism. But viruses are individual organisms under the definition, as they coevolve instead.

  14. #14 Torbj�rn Larsson, OM
    September 7, 2007

    John, on information your post is also interesting but I don’t feel it is complete.

    I believe you point out that information isn’t an especially useful characteristic or concern for biology. Likewise you note that information needs models (are semantic) to be useful.

    But information is in its technical sense slightly related to complexity as for example in Kolmogorov complexity. And as no one complexity measure can characterize all possible structures, Kolmogorov complexity is not well defined. In practice you can pick a specific compression scheme to compare information content, but you can’t do much else with that type of information.

    Therefore I feel that your objection about genetic or phenotypic information being contingent on the environment isn’t relevant by our understanding of how to measure it.

    Specifically, we can identify both Shannon and Kolmogorov information associated with genomes. It isn’t practically useful I think, but doable. Instead of giving references I will give another example.

    Apparently a model for allele frequencies in asexual populations look like Bayes theorem. And Bayes theorem is a model for trial and error inferences, where it decides which hypotheses should be weakened or strengthened. By this model we could call the populations alleles hypotheses and the population frequencies its current theory of the environment.

    This is information that the population picks up in its genome at selection by the environment. It is contingent. Changing environment will lose the meaning of the information. Over time it will make the population forget the old environment and learn about the new one.

    Again, not an especially exciting observation except for the analogies we can make. But I think we can see that even if genomes doesn’t exhibit stable and communicable information, it has definable and measurable information.

  15. #15 Torbjrn Larsson, OM
    September 7, 2007

    Um, read “static” instead of “stable” information in the last sentence. Genomes are stable and robust enough, at least for me. 😛

  16. #16 John Wilkins
    September 7, 2007

    Thorbörn: this is a basic philosophical point: the model is not the thing modelled. Yes, you can give Shannon or Kolmogorov measures for DNA strands. Yes, it is sometimes useful, just as other bioinformatics metrics are useful. No, the genes themselves do not “have” information. Our models of the genes do. There is no objective property of genes that is informational, only causal properties. Information is something that exists in our heads, not in the DNA. The “measure of complexity” depends crucially on what we choose to describe. We do not, typically, describe the information content of, say, methylation patterns – why is that? Because we aren’t interested in them, and they aren’t easily describeable in terms of discrete patterns. Yet they also affect inheritance and are “transmitted”.

    My objection is based on our liking for the text metaphor, which is a form of anthropomorphism. There is no text, recipe, or code in genes, just a process by which the primary sequence of DNA is transcribed in complex ways into the primary sequence of proteins, and even then we do not have a “coding” for function, as so much of what proteins can do depends on folding and the context in which they are employed.

    If a biological system can be well described in terms that make it an instance of an information processing system, then I will happily say it “has” information – cell-cell signalling is an instance of this, and so I can happily say there is information in the interactions between cells of, say, a nervous system. But to read information back into DNA is a feature about our own predilections, not the DNA’s. Nothing good, and a lot bad, comes of mistaking the map for the territory. For a start, we won’t find things that aren’t on the map unless we run into them headlong. The information metaphor retards progress in biological thinking.

  17. #17 Torbj�rn Larsson, OM
    September 7, 2007


    Thorbörn: this is a basic philosophical point: the model is not the thing modelled.

    I know I also run up against different term usage between, say, logic, and physics, as regards “model”. I’m conflating the use of the two types, which can be confusing.

    Our models of the genes do.

    Sure, this is the common description of information, it applies to models. It is also measurable on objects described by the models.

    Regarding your distinction between observations of objects or of relations, there is no qualitative difference in science. A quantum field can be described in a wave formalism or a particle formalism, where the wave formalism is deeper. Do you then say particles aren’t objects? And do you say that interactions between particles aren’t measurable?

    This is at the core of the difference between Caledonian’s and mine view and yours. We note that we can measure information, so it exists. Where it is localized isn’t an issue for algorithmic information theory.

    I mean that – there is no map vs territory question here. We can’t point to a specific feature of a system and say “this is the information”. What we can say is that we can observe information change. (In the Kolmogorov complexity case, difference in compression before and after. In the Shannon information case, data loss over a channel.) In the analogy I used it is by observing allele frequencies change.

    I’m sorry I didn’t make this point unequivocal – information is a relative, not absolute, measure. Personally, this is why I am skeptical to “quantum mechanics as information mechanics” ideas so far. Very much the same concern as yours, I think – it looks like a metaphor run amok.

    I tried to clarify the other differences in order to not muddy the waters. I also took care to note that these notions of information are neither necessary nor sufficient to describe evolution. I think everything you say about “reading back information” I already noted – this information isn’t communicable.

  18. #18 Thony C.
    September 7, 2007

    Caledonian wrote:

    Why are you willing to claim that there’s no definition of ‘science’, but are trying to investigate the definition of ‘life’?

    How to distinguish our host, Mr Wilkins, from myself, Thony C, a short guide for confused Caledonians.

    1) His name is John mine is Thony, now they both contain the letters “o” and “n” making a mix-up possible, however if you count the letters you will notice that his name has only four whereas mine has five.

    2) He is an Australian and lives in Australia whereas I am a Brit who lives in Germany. I know that these meetings take place in cyber-space but as both of us refer from time to time to our place of origin or place of residence this can be a strong clue as to which of us is present.

    3) He is a philosopher of biology and I am a historian of Renaissance science (mostly mathematical) and our posting tend to reflect our respective interests although he has been know to write on things historical and I make philosophical utterances from time to time.

    4) His writing style is lucid, erudite, informed, often witty and sometimes sarcastic, mines is convoluted, overblown, sometimes informative, sometimes witty and often sarcastic.

    5) He is reluctant to indulge in “slugging matches” with Caledonian whereas I seem to take a perverse pleasure in being pelted with the rotting fish that the recalcitrant Celt uses in place of arguments.

    6) He did not take place in the discussion on the definition of science/the scientific method on the Tolerance thread whereas I made several extensive contributions, although it must be said that at no point did I ever claim that there is no definition of science. In fact I pointed out that there are many possible definitions of science several of which, I will now add, are even rather useful but none of which is entirely satisfactory. The only person who appears to think that there is no definition of science is Caledonian as he has been asked more than once to state his and because he has not done so, one must conclude that he does not have one.

  19. #19 Caledonian
    September 7, 2007

    No, the genes themselves do not “have” information. Our models of the genes do.

    This is obviously wrong. Granted, it’s obvious only if you understand the technical meaning of ‘information’, and you don’t seem to like precise definitions much.

    Actually, you don’t even need to know the technical definition, as even the everyday meaning shows that claim to be obviously wrong.

  20. #20 Caledonian
    September 7, 2007

    I’m sorry I didn’t make this point unequivocal – information is a relative, not absolute, measure.

    Generally agree with your comments, Mr. Larsson, but I have to take issue with this one. The ways in which we measure information are limited and relative. The actual properties of the world aren’t relative – they’re absolute.

  21. #21 Torbj�rn Larsson, OM
    September 8, 2007


    The ways in which we measure information are limited and relative. The actual properties of the world aren’t relative – they’re absolute.

    I agree that I could have worded that much better. Especially since I tried to make an unequivocal statement. :-(

    What happened is that I wasn’t prepared to discuss metaphysics. But of course that is a risk you take on a philosophers blog. That part of the comment doesn’t reflect my new position at all.

    So I got a bit befuddled. Which I shouldn’t be because this is as you note technical distinctions on measurements, not philosophical distinctions on the meaning of life. “It’s nothing very special really.”

  22. #22 Gareth Miller
    September 11, 2007

    Trying to find the absolute meaning to the concept of “life” would be as impractical as to trying to find the absolute meaning to the concept of “good�”. What is “good” for you might not be “good” for the next person. Heck, what is “good” for you today, might not be “good” for you tomorrow. If you can’t agree with yourself, who’s to say who is right? But there is something absolute about the meaning of “good”, that is, it is the opposite of “bad”. Similarly, “life” has an opposite, and that is “death”. So the following statement is true: those who have “life” do not have “death”. Pretty obvious. Or you can say: Those who try to avoid “death” are trying to remain “alive”.

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