Skip to content

Reduction and emergence

I have been having a metaphysical debate on Twitter, which is more surreal than reading the Tractatus. My interlocutors said a few things which leads me to want to clarify my views a little more than Twitter allows. Hence…

My thesis is this:

Emergent properties are no different, ontologically speaking, than any other property. They consist solely in their surprisal value.

This is perhaps plain enough to those who know the literature, if a little simplistic. Those who do not spend their says poring over philosophical papers and books may find it a little opaque, so I will try to explain in plain language (or in technical language where I can explain that) what the issues are and what I mean by it.

Emergence is a hot topic of the past two decades. It was the subject of at least five books by that title, and each one made claims that emergent properties are critical for understanding, and explaining, the complex world of biology, society and information technology. I agree that it is crucial to understanding these things, and basically any dynamic system. What I dispute is that it explains them. To begin with, then, we have to ask, “what is emergence?”

John Holland defines it as “much coming from little” in complex adaptive systems (a phrase he coined to describe things that are, well, complex, adaptive and systems, all of which require much further definition). In Bedau and Humphreys’ edited book they note that emergence is often defined with a number of ideas: irreducibility, unpredictability, conceptual novelty, ontological novelty and supervenience (p3). It’s an interesting list. Some of these are epistemic and some are ontological. One (unpredictability) may be both. Further, the term “emergent” is sometimes applied to phenomena, to properties, and to objects and processes. Again, there is an ambiguity whether we are talking about our epistemic states (an emergent phenomenon is one we did not expect to see occur given what we know of the constituent substrate), or ontological facts (an emergent property is one that cannot be reduced under any circumstances to the properties of the constituent substrates and their relations).

Moreover, there appear to be several other claims: emergence is relative to a theory, it is the denial that physics is the most basic science, it means we cannot reduce some domains (almost always human minds, often also biology) to physics, etc. So the so-called “emergentist thesis” is a very slippery beast.

Philip Clayton has a very nice short summary of what the emergentist thesis is typically taken to be. There are four theses (taken from el-Hani and Pereira 2000): ontological physicalism is denied – what there is includes more than the physical; property emergence is asserted – genuinely novel properties emerge in complex systems; the irreducibility of the emergent is asserted – emergent properties are irreducible to, and unpredictable from, the lower-level phenomena from which they emerge (often put as “the whole is more than the sum of its parts”, also known as holism); and downward causation is asserted – higher-level entities causally affect their lower level constituents.

Each of these admits of several alternates. For example, one might not be a physicalist but still be a monist (presuming physical stuff and, say, mental alike to be different instances of some metaphysical “gunk”, to use Lewis’ term. One might allow that properties emerge novelly, but that they are still physical properties (which means, in this context, something that can be fully described in the language of an actual or ideal future physics). One might think that irreducibility is due to our inability to get the information and knowledge or theory needed, or that it is a limitation of our computational skills. Downward causation is, however, where I think the real issues lie and so I will discuss that shortly.

First, let us consider what it means to “reduce”. In biology, reductionists are akin to baby-raping robber barons. One of the worst things one can say of another in a biological context, whether in the sciences or the philosophy of the science, is that one is a reductionist. The term has been badly abused and confused. It usually means that one thinks something precious (like human cognition, behaviour or morality) can be “reduced to” genes, where “reduced to” means “is eliminated by”.

In fact, to reduce a theory to another is regarded as a good thing in science, generally. When it was first introduced by Ernest Nagel in 1961 (chapter 11), he defined it as

the explanation of a theory or set of experimental laws established in one area of inquiry, by a theory usually though not invariably formulated for some other domain. [p338]

Reduction is a case of explanatory economy, in which we make do with less in the way of theoretical apparatus to explain more than we used to. The trouble was, reduction quickly got tied into a view of domains that we can call the “layer cake” view, which is sometimes called the hierarchy view of ontology. According to this, physics supports chemistry, chemistry supports biology, biology supports psychology and sociology, etc. The idea then is that we can reduce mind to biology and thence to physics. But it doesn’t follow that we must do this. I personally think that reduction is flat: everything reduces to physics more or less directly. The layer cake is of our own construction.

This bothers some scientists: it implies that we should effectively make the discipline of biology the servant of physics. But recall the episode of Big Bang Theory in which this conversation occurs:

Penny: [surprised that Sheldon actually might actually help her with her flower beret business] …and you know about that stuff?

Sheldon: Penny, I am a physicist. I have a working knowledge of the entire Universe and everything it contains!

Penny: [sarcastic] Who’s Radiohead?

Sheldon: [suddenly his eye begins twitching, because he’s frustrated he doesn’t know the answer] I have a working knowledge of the important things in the Universe. Good luck!

Physics tells you what may happen in the universe. It will not tell you what actually does, unless you go out and look at the boundary conditions. Anything that happens, if the universe is simply physical, must be explicable in terms of physics but what physics actually obtains is a matter of investigation. Physicists know nothing until someone finds out what the boundary conditions are, so stamp collecting is not only a valuable enterprise, it is crucial to knowledge.

So a science like biology or psychology can be autonomous to some degree. But the history of science shows that physics, and its daughter chemistry, tends to creep into all manner of domains. Much of biology now is physics and chemistry, although it goes by the names protein chemistry, molecular biology, biochemistry, and biodynamics, and so forth.

Moreover, the autonomy of these disciplines is either to be understood as ontological (there are things in these disciplines that do not have any existence in physics) or epistemic (we have to name and infer using entities of the discipline because we cannot make the inferences from physics to, say, biology). So let us take biology as the target domain, and ask: is the autonomy of biology ontological or epistemic? The question can be reframed to cover mind, society, or information technology, etc., as it suits you.

It’s pretty clear that biology is autonomous from physics to some degree, but the historical trend has been that this autonomy is not constant, but is decreasing. Historically, biology began as its own field largely because we understood so little about how, for example, organised bodies (from which the term organism derives) do things in physical terms. As we learned about things like the Krebs Cycle, cell walls and membrane structures, and the genetic code, to mention only the most widely known aspects of living systems, we began to see how living things were simply physics organised in particular ways with particular dynamics. It took us a long time, and we are not entirely free of it yet, but we learned that there is nothing “vital” about living systems that is not physical.

More and more things were reduced in biology. Sometimes they were reduced to chemical things (and no further, because the chemistry was either not understood or the details were irrelevant to the biological explanation), sometimes to physical things, and sometimes (as in the Lotka-Volterra equations of predator-prey cycles) to mathematical items and models.

Assuming that this will continue (question begging, I know, but bear with me), biology is not ontologically autonomous. If we can explain everything without remainder, there is nothing specifically biological that is not also just physics arranged in a particular way. So there is an inductive argument that reduction is correct. What weight you think this carries will depend on what value you place on inductive arguments.

So autonomy is something that can merely reflect history and our cognitive limitations, and no strong conclusions can be drawn from it.

And yet, every scientist working in sciences of complexity will affirm that there are properties that emerge in surprising ways from the underlying physical substrate. I cannot deny this on pain of being silly (this is the argumentum ab silliness, err, reductio ad absurdum). So we need to account for this if we are clubfooted reductionists* like I am.

To deal with this, I appeal to an example of John Stuart Mill’s, and a notion from information theory – surprisal value. In the next post…

*I owe this phrase to Paul Griffiths, who used it in passing. I have since adopted it as a badge of honour.


Bedau, Mark, and Paul Humphreys. 2008. Emergence: contemporary readings in philosophy and science. Cambridge, MA: MIT Press.

Clayton, Philip. 2006. Conceptual foundations of emergence theory. In The re-emergence of emergence: The emergentist hypothesis from science to religion, edited by P. Clayton and P. Davies. New York, Oxford: Oxford University Press:1-31.

El-Hani, Charbel Niño, and Alfredo M. Pereira. 2000. Higher-level descriptions: why should we preserve them. In Downward causation: Minds, bodies and matter, edited by P. B. Andersen. Aarhus: Aarhus Univ. Press:118-142.

Nagel, Ernst. 1961. The Structure of Science: Problems in the Logic of Scientific Explanation. London: Routledge and Kegan Paul.


  1. Jim Thomerson Jim Thomerson

    I have understood emergent properties to be the result of organization of components of a system. Reduction is study of the components. Holism is studying the complete system in action. Both are necessary to understanding the system. Consider your automobile engine. What can be learned about it while it is running? What can be learned about it reduced to component parts on the workbench? Why is it taken for a test drive after repair of the components?

    • It is often claimed that relations between parts are not covered by reductionist accounts – a claim I find absurd and strawmannish. Of course we have to take relations (that is, the organisational networks) into account. I would find it amazing had anyone ever said anything else. I will deal with this in the third post.

  2. Adam Adam

    My biochemistry textbook in college had a quote from Albert Szent-Gyorgyi, that ended with the sentiment that “In my search for the secret of life, I ended up with atoms and electrons, which have no life at all. Somewhere along the line, life has run out through my fingers. ”

    My understanding of “reductionism vs holism” is similar to that described by Jim Thomerson, above. So I suppose that this is “epistemic” holism (if I’m using that term correctly).

    This sort of holism may seem trivial, but my impression is that many theories (and research programs) have been driven by a misguided reductionism … where particular components of a system were taken to embody the essence of the whole.

    For instance, there was vitalism, which supposedly proposed that the essence of life was some “vital” component, rather than being a property of the entire system. Likewise, I’ve heard that some primitive neuroscientists thought that “the soul” had residence in a particular part of the brain. So there is a misguided type of reductionism, which examines components in isolation, as though these isolated parts carry the properties of the whole, when in fact, it is the interaction of the parts that produces the properties of the whole. If someone were to say that human cognition could be reduced to genes, he would be making exactly this type of error… cognition requires not only genes, but the cells that the genes are embodied in, and the entire brain, and the entire body, and in many respects, human society; genes are only part of the process.

    Many arguments against “reductionism” are not arguments that we should ignore the details of a system, but simply that those details need to examined in the context of the entire system. There are also cases where the details are a distraction (once we demonstrate that basic physical laws are accounted for), and it is important for us to recognize when that happens.

    • Adam Adam

      I just want to acknowledge that while writing the above comment, I had missed a major point in the essay because I misunderstood the term “reductionism”.

      Also, I wasn’t disagreeing with any major conclusion of the essay… mainly I was responding to the implication (in retrospect, one that was probably not meant to be implied) that biologists tend to be bleary-eyed mystics — even though physical theories such as thermodynamics can be found everywhere in biology from biochemistry to ecology. Even Lovelock’s “Gia theory” is reductionist in this sense.

      I also chafe at the idea that biology is just a special case of physics… as though all this work would be irrelevant if we simply had good enough computers and good enough measurements.

  3. Adam Adam

    “sometimes to physical things, and sometimes (as in the Lotka-Volterra equations of predator-prey cycles) to mathematical items and models.”

    I don’t understand how something can be reduced to “mathematical items and models”. I’d think that in this case, population processes were reduced to the behaviour of individual animals.

    • Adam: I don’t understand how something can be reduced to “mathematical items and models”.

      I had the opposite complaint: how can you not reduce something except using mathematical models? (The “chemical things” and “physical things” are mathematical models as well.)

      But to answer your question: The Ideal Gas Law reduces atoms to identical, countable particles that can be wholly described using their population (N) and a single coefficient (k), thus PV=NkT.

      The Krebs Cycle is far more involved, but “chemical equations” are no less mathematical for being in a different domain.

      I’m not going to say that one can’t create non-mathematical scientific models, but I sure am having difficulty bringing a modern example to mind.

      • Adam Adam

        I don’t object to the USE of mathematics as the language of reduction, I just object to saying that it has been reduced TO mathematics.

        If anything can be reduced to mathematics, then everything is reduced to mathematics, and the main point of this essay (that theories are reduced to physics) is undermined.

        I’m sure that some people who believe that mathematical equations are real in the same sense that molecules are real (I don’t), but everything in the above essay implies that the process of reduction involves only empirical sciences. Therefore, math would be excluded from this process (except as an accessory).

      • Adam Adam

        Regarding the above examples…

        I think that the ideal gas law is comparable to the Lotka-Volterra equations, in that both describe how the properties/dynamics of a population arise from behavior of individuals within the population (assuming a homogenous population).

  4. Markk Markk

    I always thought the key point of emergence was independence. That is I have a model of turbulent fluid motion. I have another model of classical momentum and energy conservation. I have yet another model of Quantum Mechanics. For the purposes of my first model, I generally can substitute in either of the other two and things don’t change a bit in the domain of the first model. It does change as I try to create the first model as a set of actions in either of the other two models, but from the point of view inside my turbulence model it doesn’t matter – it has emerged from the other two and I can make correct predictions within its range of use. It only needs to be consistent with those other models. In fact once could say that once these predictions are confirmed, to that degree they become constraints on the second two models also. I think generally we call these constraints =confirmations= when they are consistent.

    So emergence to me only relates to scientific models. It has nothing to do with reality itself. It has to do with the independence of one model from another and the fact that rules of one model do not depend on another model to the degree that other model could be changed out with only some properties still applicable.

    • This is a crucial point to make and one I shall return to. Since I think emergence just is epistemic, the independence of models is precisely what emergence consists in. We then must explain why the models are independent, a topic covered in the philosophy of science under the rubric “the unity of science”.

  5. bob koepp bob koepp

    “Physics tells you what may happen in the universe. It will not tell you what actually does, unless you go out and look at the boundary conditions.”

    I realize this is peripheral to your main focus on reduction and emergence, but I must quibble with this formulation. Suppose we accept, for the sake of argument, that F=ma. Then, this bit of physics tells you what may happen in the sense that “whatever is physically possible must be consistent with F=ma.” But further, it tells you that if cases of acceleration that actually occur are in apparent violation of this equivalence, then there must be forces and/or masses that have not been properly accounted for in our accounting of boundary/initial conditions. In other words, if F=ma is true, then it does indeed tell you what actually happens in the universe.

    • It’s a conditional knowledge – *if* there are objects in that universe, then f=ma describes their dynamics. But for all the (Newtonian) laws of physics know, the universe (or that part of it) is empty. I dub this the Radiohead conundrum.

  6. Jeb Jeb

    I often go through periods of wondering what the hell I am doing reading a philosophy of science blog and then posts like this come along.

    “Emergent properties are no different, ontologically speaking, than any other property. They consist solely in their surprisal value.”

    I’ve never come across the term emergence, I must be picking it up second-hand from somewhere as I appear to be using similar ideas, although I can’t express them with the same clarity or pin point where I’ve picked them up.

    I’ve just spent the last week trying to come up with the words to resolve a problem in a very different subject area. Two sentences seem to hit it full in the face, directly with a simplicity and clarity that is always nice to see.

    It seems to cut straight through to the core of an issue that has been bothering me of late.

    ‘Surprisal value’ is particularly nice.

    I must use and cite this a.s.a.p.

  7. Adam Adam

    For what it’s worth, I just came across an example of what I’d call “mystical emergence” in this review of “Aping Mankind”

    “Rather what he objects to… is the abuse of science. In this case, it means the attempt to reduce humanity to little more than unthinking matter, as utterly subject to the same laws of evolution, indeed of physics, as every other material object, organic or otherwise.”

    “For the stuff of consciousness does not belong to the quantifiable and abstract realm of the physical sciences, it belongs to the qualitative and phenomenal world. As Tallis writes: ‘The very notion of a complete account of the world in physical terms is of a world without appearance and hence a world without consciousness.’”

    “our actions are so ‘irreducibly complex’ that the notion of a cause becomes inapplicable. What is the cause, for instance, of my reading Aping Mankind? Tallis’s account of freedom, his reassertion of humanity, does not rest on an alternative account of a self-willed causal chain. Instead its better grasped, as he himself put it, as an attempt to ‘wrest the truth of time from physics, because the truth of time is about anticipation, regret, nostalgia, joy, hope and so on.’”

    I’ll make one note about the sloppy scientific reasoning of this review: the reviewer acts as though all the evidence of a brain-mind connection is observational/correlational. At no point did I see any acknowledgement that physical and chemical treatments can radically alter “the mind”.

  8. Bill Morse Bill Morse

    My first reaction was that life is in fact ontological wrt physics – the idea of being alive has no meaning in physics. That is, physics can illuminate life, but nothing in physics can be said to be alive, and so biological questions are by definition the domain of biology.

    But on further reading I don’t think that is your argument, so instead I will take issue with your statement that ” there is nothing … about living systems that is not physical.” (I deliberately snipped the vitalism reference)

    One of the classic examples of quantum mechanics is Schrodinger’s cat. But that is rather arcane, so let us elaborate on this a bit and bring it into the everyday world. If we are reducing life to physics, what determines whether an organism lives or dies?. OK in the case of massive trauma we know all the systems are stopped. But what about the organism that is on the edge? Surely by your reductionism this has to come down to a single electron that when in one state dictates the organism will live but when in another state determines the animal will die. And yet we know from physics that we cannot predict the state of that electron. So physics cannot tell us whether an organism will live. Thus physics cannot ultimately encompass biology.

    I have additional arguments that pertain to arguments brought up in “Reduction and Emergence” but they seem more germane to some of your later bloggings so I will respond there.

Comments are closed.