Downward Causation

The final claim for there being an ontological sense to emergence is “downward causation“, a phrase coined by the evolutionary epistemologist Donald Campbell in the 1970s. The idea here is that emergence is real because higher-level (or bigger, composite) entities cause changes in the properties and dynamics of their parts. This is very big in the philosophy of mind. It is also often used by a certain kind of systems thinking in physics. Recently, Sean Carroll took it to task.

This view requires a mode of thinking sometimes called “hierarchy theory”. It is roughly this: the world is comprised of objects organised in a hierarchy: subatomic particles form atoms, atoms form molecules, molecules form monomers, monomers form polymers, polymers form cells, cells form organs, organs form organisms, organisms form societies, societies form economies, etc. It’s a very neat way of conceiving of the world. And, if not actually always false, it’s usually false. Some molecules only form molecules (molecular water is far more common than fluid water, although not on earth). Some cells, well actually the bulk of cells, form single celled populations and little more, although many form biofilm communities. And so on.

Hierarchy theory was a needed correction at a time when reductionists were being atomistic, back in the 60s. But like any spice, as Kierkegaard remarked, you cannot make a meal of it. And even if some things are arranged in mereological hierarchies, it doesn’t follow there is only one such hierarchy. For example, there are protein complexes both inside and outside cells. I do thinking with both my little grey cells (as Poirot would call them) and also my iPhone, Mac and Google. Hierarchy theory is largely an echo of the Great Chain of Being, but that’s another post for another day.

So what does the claim there is downwards causation consist of? A simple formulation is by Bernd-Olaf Küppers: “The whole determines the behaviour of its parts” (Küppers 1992:243). That is to say, the causal arrow points from larger or organised whole to part. We can diagram these:


Causation occurs at higher, lower and intra levels. So if the higher level is mind, and the lower level is brain, then mental events can change the behaviours of the brain, as well as vice versa and between level events. By contrast, the reductionist account has causation at some lower level only:


On at least one account, higher level causes aren’t really causes at all, but just relations between what are themselves just compositional objects and apparent processes. A mental even cannot cause brain processes to change behaviour because it has no causal power that is not simply a composition of the underlying (physical) processes. What looks like downward causation is just lower level causation of both the whole and the part. If my fear causes me to express more adrenaline, it is because both fear and adrenaline production are physiological (and hence physical) processes.

Notice that I did not say “just” physical. If everything is physical, then dismissing the phenomena as illusory is unnecessary and prejudicial. A mountain may be composed of rocks and ultimately subatomic particles and fields, but understanding that doesn’t mean I don’t have to go around or over it to get to the other side.

Emergentism of this kind (there are several) is unlovely, philosophically, as it has way more causal arrows than are needed. Or does it? Our intuitions are that what happens at the higher level do in fact influence what happens at the lower, and so things like “temperature” causes molecules to move about more rapidly. I think what happens here is that an explanation moves downwards, but only because we have hypostatised (thing-ified) an abstract term. It’s very little more than definition. But causation, if it occurs at all, occurs at the “lowest”, which is to say, the physical, “level”.

So now the question is what that level is. As a physicalist, I of course think it is physical, but since our best physics is not yet right, I can’t say that it is actually at the level of the particle zoo as now understood, and so on. It doesn’t really matter, though. I presume (thus making a pres out of u and me) that we will one day unify physics into a universal theory. I doubt greatly that it will turn out to be all that different to our best modern physics. But even if it were, the question concerning emergentism is whether or not we can reduce other theories to other theories than physics nearly all the time. We do not attempt to reduce psychology to physics (yet!) but to biology. So whether or not we unify physics, now or later, we can still usefully be reductionists.

I said before that to be a reductionist in biology is akin to being a baby-eating robber baron. This particularly played out in the sociobiology debates of the 70s. It need not have. While we may object to genetic determinism, we can hardly equate that with biological determinism. I can think (indeed all the evidence points there) that we are what we are because our biology makes us and still allow for environmental determinism because biology always has a context. But be determinists we must: it is identical in intent if not terminology to saying that things have natures, which is the foundational view of science itself. We act according to our natures, just as God is supposed to, and most of our moral theorists have argued for two thousand years. To think otherwise is to deny the possibility of knowledge.

I know that people want to preserve our specialness, especially our mental specialness, but if wishes were horses then beggars would ride. The world is not required to match our desires, despite what Oprah would have you believe. An ancient doctrine (“as above, so below”) of the alchemists had it that we do what we do because we mirror the intentionality of the universe – in fact it is the precise reverse. We presume the world is like us. But we are like the world, and just as unintended and as determined as everything else.

Reductionism, or rather than variety of it known as physicalism, makes the best case for understanding the world, including ourselves. Emergentism of the strong, ontological, variety is mysticism. Wholes are more than the sum of their parts precisely because they have structure that the parts, by definition, cannot have. Allow reduction to include spatial, causal and temporal relations, and you get everything you need. You just don’t get that specialness we so ardently desire. And so what?

The virtue of post-Baconian science is that final causes were abandoned, except (and this is often overlooked in the rhetorical flourishes of this old debate) where you have a system capable of projecting goals and seeking them. And these systems (which you may gather includes humans) also call for an explanation: how can systems seek goals? The answer is fairly simple. They do so because their ancestral systems did so, only perhaps less well. There is nothing special about these systems in an ontological sense – they are just composites of less complex systems, and have degrees of freedom that the parts don’t have. A wheel will not win the Indianapolis 500; a car can. That doesn’t mean “car-ness” is a novel ontological category that causes wheels to turn and follow race tracks.

I have rambled somewhat, largely because I don’t intend to publish this, but to stimulate debate. So, you strong emergentists, make your case. Why should we think that emergence is anything other than the point at which our expectations and computational capacities are exceeded? Why believe in downward causation? Don’t trot out the old intuitions and examples, but defend yourselves. For now, I see no need to move beyond physics. In the end science will reduce everything directly to physics – no layer cakes here. We have warrant for this, based on the past. Sure, we may one day meet something that can never be so reduced, but we haven’t yet.

This also goes to the question I asked a while back: what are the domains and disciplines of science? I will ask this again in a new post.


Beckermann, Ansgar, Hans Flohr, and Jaegwon Kim. 1992. Emergence or reduction? Essays on the prospects of nonreductive physicalism. Library ed, Foundations of communication and cognition = Grundlagen der Kommunikation und Kognition. Berlin; New York: W. de Gruyter.

Campbell, Donald T. 1974. “Downward causation” in hierarchically organized biological systems. In Studies in the philosophy of biology, edited by F. J. Ayala and T. Dobzhansky. London: Macmillan:179-186.

Küppers, Bernd-Olaf. 1992. Understanding Complexity. In Beckermann et al. :241-256.


24 thoughts on “Downward Causation

  1. I can think (indeed all the evidence points there) that we are what we are because our biology makes us and still allow for environmental determinism because biology always has a context. But be determinists we must …

    You use the term “think” a lot here. What is its definition/nature in a determinist/physicalist context?

    1. It’s a dynamic of a physical system that processes symbols, and dispositionally emits informational outputs. The system has some syntax, but mostly just processes inputs and outputs in a neural network that is embedded in a larger physical system. which includes the environment of other similar physical systems. The use of symbolism and syntax leads a lot of these systems to think there is some reality to language beyond the uses to which these systems put it.

      1. So “processing symbols” and “informational outputs” can be reduced to physics? What purpose is there to the symbol “think” then? Why not just say “I am deterministically disposed to process symbols in such a way as to informationally output a defense of physicalism”? Sure, its clumsy, but since most people … um … think that thinking has, at least, some volitional content, wouldn’t it be clearer to state that you have no choice but to say what you said?

        1. But we obviously do have a choice (volition) even though we are determined. This is the standard compatibilist position. It’s silly to suppose that finding out that we’re deterministic systems would somehow mean that we’ve discovered that we don’t think.

  2. So if we take this view as it is stated, how would one formulate a reply to Searle when he talks about consciousness as being ontologically irreducible. Could one give a satisfying account of consciousness without falling into the allure of non-reductive physicalism?

    1. I don’t know what Searle can mean by the term “consciousness” if it is non-reducible. How does he know what it is if it is not explicable and approachable by ordinary empirical means? It’s a meaningless term unless you break it down to simpler things like feedback loops of control systems.

      1. That seems an extraordinarily odd argument. Since reductions are forms of explanation, you can’t have them unless you already have their explananda; which means that knowing the explananda has to precede any analysis of it into simpler things. It is true of nothing else whatsoever that the term for it is meaningless if we lack a reduction of the phenomenon into simpler things; consciousness may not be special in other ways, but there is certainly no reason to think consciousness is magically different on this point.

      2. Well, Searle talks about consciousness as being causally reducible, but not ontologically reducible. I’m not sure how that works exactly.

      3. It’s a meaningless term unless you break it down to simpler things

        This can’t be a general rule, since it would make the goal of science to reduce everything to the meaningless. I tend to follow William James on the idea of “consciousness”, so I don’t know that I can follow the objections laid out here, but it does seem that “science” – as a concrete body of knowledge that a particular person knows – is a sort of whole that exercises the causality of making certain parts meaningful or meaningless. The ability to judge particular things in light of general rules seems is hard to separate from a top-down causality, and any attempt to divide the noetic order from the world- order too sharply will leave us with dualism. I often get the sense that you unintentionally allow this sort of of dualism when you make such an extreme break between “when we’re talking about things, and when we’re talking about words”. If the domains are really as divided as that, then the noetic domain has rules that are different in kind from the world domain, and pluralism in rules makes for pluralism in sciences.

  3. While we may object to genetic determinism, we can hardly equate that with biological determinism. I can think (indeed all the evidence points there) that we are what we are because our biology makes us and still allow for environmental determinism because biology always has a context. But be determinists we must: it is identical in intent if not terminology to saying that things have natures, which is the foundational view of science itself. We act according to our natures, just as God is supposed to, and most of our moral theorists have argued for two thousand years. To think otherwise is to deny the possibility of knowledge.

    Isn’t determinism usually something much, much stronger than this? Roughly: given the state of the system is in at time t (and the laws or whatever for the changes in the system over time), for any other time t’ there is only one state that the system can possibly be in at t’.

    Borrowing from Nancy Cartwright a bit, it seems plausible to me to say that a thing has a nature if, and only if, there are true ceteris paribus generalizations about the changes in the state of the thing: ceteris paribus, if I put my Bolivian rainbow pepper plant outdoors for the summer and make sure it has plenty of water, then by season’s end it will have grown several inches and produced a couple dozen little red (and wonderfully hot) peppers. My plant acts according to its nature, and I use this to my advantage, growing peppers to use for cooking in the winter.

    But, of course, things could go wrong with my plant. For some reason or another, it could end up under- or overwatered. Or aphids might kill it. Or it could get some disease. Or its pot could be broken, exposing the roots and killing the plant. And so on. So, given its state now (flourishing on my front porch in the morning sun), there are almost no other times t’ such that there is only one state that it can possibly be in at t’. My plant has a nature, but it’s not strictly determined by its nature.

  4. Note that one might plausibly believe, on a slightly different variety of reductionism, that, according to one’s definition of “cause,” one might have causal arrows (and not just epiphenomenal arrows) pointing between the higher-level things but still not across levels. I find this a particularly attractive way of thinking about the philosophy of biology: there’s one causal story we can tell where organisms are nothing but causally efficacious clouds of micro-particles, and there’s another story we can tell whereby they’re organisms that do things in environments. The second story’s not incompatible with the first, because the second story just is a clever way of summarizing the facts found in the first story. No less causally potent (the tarantula hawk really does cause the death of the tarantula), but not “spooky” or “emergent” or “downwardly caused” or what-have-you. Just so happens that in our world, clouds of small things gang up in ways that make them into causally potent big things.

  5. It certainly appears that the explanation for the existence of so many books about the mind-body problem is the puzzlement people feel about their consciousness of things. It looks like a higher-order phenomenon is causing a lot of lower order ones. Insisting that puzzlement about consciousness is an error doesn’t dissolve this problem. It just means that one has to explain how an illusion can have causal force even though illusions are not obviously made of atoms and void even if deluded individuals are. One can imagine a true account of the production of Consciousness Explained that reduced the process to a heck of a big collection of Feynman diagrams, but that planet-sized explanation would still apparently leave something out, i.e. the obvious reason Dennett wrote the book.

    I’m not presenting this case as a counter example because I don’t know if it creates any problems for your version of reductionism. It’s just that I’m not sure how you would handle it or the somewhat similar issue of how an account of the phenomenon of pain that only involves the agitation of atoms finesses the fact that pain hurts and is avoided for that reason.

  6. John, when you’re right, you’re right. And you’re right.

    Two minor quibbles: (1) There actually are strongly emergent properties (in which the property of the whole is more than the structured sum of its parts) in quantum mechanics. The properties of entangled states cannot be ontologically reduced to the properties of the components (unless you buy into instantaneous action at a distance). But such holism is restricted to the quantum level; it’s irrelevant to the general metaphysics of weak emergence that you’re defending here.

    (2) I’m not sure you get the metaphysics of degrees of freedom of composite systems quite right (perhaps in part due to a typo?). You say:

    they are just composites of less complex systems than have degrees of freedom that the parts don’t have.

    But the full system — in its microphysical characterization — will always have more degrees of freedom than does a composite system. The wheel can’t win the the Indy 500, but it’s a mistake to just consider a single part. You have to consider all the parts that compose the system. And then the relevant point is that the parts can always do more than the whole can. What makes something a “whole” is that you rule out certain behaviors of the parts and thereby reduce the number of degrees of freedom. So the parts (as a group) do have the degrees of freedom of the whole, and they have extra degrees of freedom as well.

  7. If Physicalism means that (a complete) physical science will explain all the facts of human experience, one objection to it has the structure:

    ______ is a fact given in human experience.
    Physical science never treats of _____ as such

    Some candidates to fill in the blank are:

    1.) A first person experience
    2.) Ineffable experiences (religious experience/ psychedelic hallucination/ rapturous)
    3.) Non-univocal / generic experience
    4.) The particular thing in its concrete particularity
    5.) The interpersonal
    6.) The non-metrical
    7.) The qualitative
    8.) The historical
    9.) The Euclidean
    10.) The transcendental (the various distinctions among being as being)
    11.) The prudential

    Physical science can’t treat some of these at all, and it can’t treat any of them as such, since in treating of them it translates them into a different sort of experience. This translation does not commit science to the idea that it has captured the one and only sense of what things are, just as someone who uses a pen to write down what someone says is not committed to the idea that speech is really made of ink.

  8. Perhaps all that reductionism really means, is “explaining something in terms of something else”. For example, I can explain the origin of my existence in molecular terms, evolutionary terms, personal terms, cultural terms, astrophysical terms, or (if I was religious) in theological terms. Size is not a common factor in any of these explanations. But none of these explanations could be properly classified as causal, whatever that means. They merely seek to complete an established pattern of consistent observations. One must also realize that all observations actually do occur at the level of human senses. I do not observe an atom, but the display from a machine, and the print from a science book.

  9. Quite a lot of interesting and intellectually challenging comments on this one – you appear to have struck a nerve 🙂

    To add my two cents:

    I don’t know that I am a strong emergentist, but in any case I think your discussion of top-down vs. bottom-up causation ignores the elephant in the room, which is feedback. AFAIK most phenomena claimed to be emergent involve feedback. And where is the causation when feedback occurs? Is it bottom up? Well on the beginning of the first loop it is. But how about at the top of the loop? Then causation is top down. But wait a minute – when we start the second iteration of the loop, the causation is bottom-up again. Et cetera, et cetera. There is a good reason why Hofstadter pays so much attention to feedback, because it is key to understanding how complexity arises. But it does not comport to our standard ideas about causation.

    1. Feedback is not a special kind of causation – just the ordinary kind that has a particular network topology. So it is not a counter-instance to my rabid reductionism, because all causation is still just physical. What happens at “higher” levels is still constituted by the component parts and their physical properties.

  10. Late addition… because I saw a quote from this blog in a recent comment on your Feb 28 blog.

    You say: [i]Wholes are more than the sum of their parts precisely because they have structure that the parts, by definition, cannot have. Allow reduction to include spatial, causal and temporal relations, and you get everything you need. [/i]

    Clearly put! I coined another way of saying this for myself (I think; unless I recalled it sunconsciously).

    [i]Wholes are not the sum of their parts. They are the product of their parts.[/i]

Leave a Reply