Monday, March 8, 2010

Is Physics Ontologically Basic?

Last week, I attended a talk by a fellow contributor to this blog -- I don't think it's good to identify her or him since I want to discuss a view I only heard.  I may easily have misunderstood the view, and the view I state may or may not be the same as a view that has been published by the speaker, or may distort what the speaker wanted to make of it.

So let's call the speaker X: if X reads this and wants to correct me with or without self-identification, so much the better.

X argued against the "reductionist" idea that some sciences are more basic than others.

1.  All scientific theories are representations of reality, and as such they are incomplete since any representation leaves out something.

2.  Let R1 and R2 be distinct representations of reality, neither of which is translatable into the other (e.g., by bridge principles).  Suppose further that there is no representation R3, such that R1 and R2 are both translatable into R3.  Under such circumstances, there is no basis for saying that R1 is more basic than R2, or vice versa.

3.  Since there are no bridge principles, any two sciences (e.g., physics and chemistry; neuroscience and psychology; ecology and biology) are like R1 and R2 above.

4.  Therefore, no science is more basic than any other.  Any two sciences are different incomplete representations of reality -- each gets at features of reality that the other leaves out.

Now, it seems to me that this is a good argument for many conclusions.  For example, it is a good argument for the autonomy of chemistry, psychology, ecology and so on.  Of course, this depends on establishing premise 3 for the particular case -- but let's just grant that this can be done.

Now in the discussion of X's paper, some people urged the following sort of "objection" -- it isn't really an objection; more like an independent point.  They said: Look, let's grant the above argument, but couldn't it still be the case that the reality that physics investigates is ontologically prior to the reality that chemistry investigates? -- perhaps in the sense that chemistry-facts are supervenient on physics-facts, or that chemistry-facts are just physics-facts under certain combinations, or something of the sort.

X's reply, or at least the reply I thought I heard was this: well, I agree that chemistry-entities are made up out of physics-entities, but because there are no bridge principles, the supervenience claim cannot be true.  After all, there is no science in which it is true.

I am unsatisfied by this.  I am inclined to think that IF chemistry-entities are made up out of physics-entities, then supervenience holds.  (I don't want to put too much emphasis on supervenience: the claim that every chemistry-fact is a physics-fact is good enough for me -- some "ontological" claim of this sort.)

Now, the supervenience (or inclusion) claim may be of no interest to the chemist, since s/he has autonomous methods of investigating the domain.  Further, it may be of no interest to the physicist, since it is messy and, moreover, reveals very little of interest concerning the physics-domain.  It may even be the case that there is no effective way of stating the supervenience relation.  But this does not imply that the supervenience claim is false.

What's wrong with my intuition?


  1. If you believe that

    "chemistry-entities are made up out of physics-entities"

    and also

    "there are no bridge principles"

    I'd want to know what you mean by 'made up out of'. It looks like you have a bridge by saying that one is 'made up out of' the other, but there are also no bridges.

  2. What's a "bridge principle" when it's at home?

  3. What does "ontologically prior" mean? Is that a temporal claim, as if a hydrogen atom (the "physics entity") exists BEFORE the element of hydrogen (the chemistry entity) exists? Or is there some other sense of "prior" here?

    Do atoms exist independently of elements? How does that work?

  4. Good questions.

    To begin, a bridge principle is one that tells you which physics-entity a given chemistry-entity is -- for instance, that a hydrogen atom is such-and-such fundamental particles in thus-and-such relations.

    "x is ontologically prior to y" means that ys are nothing over and above xs in certain relations -- for example, that chemistry-entities are nothing over and above physics-entities in certain relations.

    Finally, Noah, I share what I take to be your perplexity: if hydrogen atoms are made up of fundamental particles, then why wouldn't there be a bridge principle? (Remember: I was reporting a view I don't understand.) So if a hydrogen atom is just these fundamental particles in thus-and-such relations, why isn't the behaviour of the hydrogen atom just the behaviour of said fundamental particles? And if physics is complete, why isn't it possible in principle to predict the behaviour of hydrogen in physics (under the description given).

  5. It seems like we have a differential burden of proof in operation. Why assume "chemistry-entities are nothing over and above physics-entities in certain relations?" Why aren't physics entities simply chemistry entities absent certain relations? And what kind of ontological work is being done by the phrase "certain relations" anyway?

    Let me ask the question this way: If we have a a single hydrogen atom, do we have the element of hydrogen? Those who make this claim that physics is prior to chemistry must establish that a single hydrogen atom IS NOT the element of hydrogen. If they do not do so, then can't a chemist simply claim that physics is nothing but chemistry in little tiny bits.

  6. Supervenience requires not only material composition, but also causal/behavioral composition. It seems that X's view could then be that chemical-things are made up of physical-things, but the behavior of chemical things cannot be explained (or better, is not only governed by) the laws of physics acting on their physical parts. This is similar to older arguments for emergence, plus burden-shifting.

    Strictly speaking, though, X would be on better ground if they claimed instead, "because there are no bridge principles, there is no reason to believe that the supervenience claim is true. After all, there is no science in which it is true." Perhaps this could be strengthened with a commitment to methodological naturalism.

    An important problem with X's view might be: chemical-things are made of physical-things, but physical-things aren't made up of anything--this asymmetry implies ontological priority of the physical. Response: Pick your favorite thing at the other end of an ontological hierarchy (universes and societies both seem good candidates). Those things are composed of other things (astronomical objects; persons and institutions) but are not themselves parts of larger things. This asymmetry implies ontological priority (of the whole universe, of the social).

  7. >> I am inclined to think that IF chemistry-entities are made up out of physics-entities, then supervenience holds.

    I'm inclined to agree with you, Mohan. And the reason is, our evidence for the antecedent comes from our scientific theories -- in particular, theories with reduction relations (or "bridge principles" or whatever) between them! Suffice it to say, I'm very puzzled by Premise 3 in this argument. Energy, equilibrium, bonding, and phase are all concepts in chemistry that can be reduced to concepts in physical theories. What's the problem?

    Of course a complete reduction of chemistry to physics may not be possible -- but the existence of any partial reduction principle is enough to refute Premise 3. How did X defend this sweeping claim?

  8. Two points, jpj:

    Generally, atomists assume that the behaviour of wholes is determined by the behaviour of their parts, or of their independently existing parts. Of course, this is controversial -- people have argued for "downward causation", in which the wholes influence the behaviour of parts.

    But the principle you suggest: namely that the behaviour of parts is determined by the behaviour of wholes of which they are parts is even more controversial -- especially if "determined" is taken to be (a) complete and (b) a directional causal relation different from (for instance) "predicts", which can be symmetric.

    Can you think of ways that the behaviour of hydrogen atoms causally determines the behaviour of the fundamental particles that make up the hydrogen atoms? That's generally not thought to be the direction of causality.

    A second, smaller point: physics entities can't be chemistry-entities without relations because (presumably) the physics-entities could exist independently of being part of chemistry-entities, but not vice versa.

  9. Matthew, your point that material composition is not enough for supervenience is a good one. Still, an argument is needed to persuade me that chemistry-entities are materially but not behaviourally composed of physics-entities. I don't think that the paucity of bridge principles is going to do the job. As Bryan says: "Energy, equilibrium, bonding, and phase are all concepts in chemistry that can be reduced to concepts in physical theories. What's the problem?" (Though I know some philosophers of chemistry who would dissent.)

    Bryan, good question about bridge principles. The arguments I understand show only either that bridge principles are not law-like in themselves because higher-level entities are multirealizable, or that they are too cumbersome. However, there are some highly technical arguments given by John Earman and others that suggest something more interesting, namely that the interactions between physical entities in chemical complexes are not physically determined. I don't think that these arguments generalize to say the relationship between biology and chemistry.

    Finally, Matthew: I don't think that the top-end asymmetry you mention will establish ontological priority if the composition principles work.

  10. Could the problem simply be that you don't think that you can draw metaphysical conclusions from epistemological premises? The supervenience claim is clearly a metaphysical one... how could it be refuted by the plainly epistemological point that "there is no science under which it is true"?

  11. You've summarized my perplexity nicely, Nick. But X claimed not to understand the metaphysical/epistemological distinction. (I think that's a way of saying it's incomprehensible.)

  12. Mohan, I think it all hangs on how X defends skepticism about bridge principles, and whether you then accept the burden-shift. If you think the burden is on the anti-reductionist to explain how the emergent behaviors emerge, then X is in a tough spot. If you think the burden is on the reductionist to make plausible that supervenience holds, then the point about bridge principles seems apropos.

    The top-end rebuttal of course only works if you're trying to argue for ontological priority on the basis of material composition without supervenience holding.

  13. Matthew, X could probably dispense with talk about bridge principles. How about making a distinction like this: non-fundamental entities are those whose behaviour is determined by their components; fundamental entities are those that display at least some behaviour that is not determined by the behaviour of their components. Then the question is: Are there fundamental entities that are not physics-entities? This is the emergence question. As you pointed out earlier, the answer to this question is not settled by the concession that everything, regardless of domain, is made of physics-entities.

    A second question is: Are there FUNDAMENTAL entities that influence the behaviour of their components? (Shapiro and Sober say that there are causal links from composed entities to their components, but they acknowledge that the behaviour of the composed entities supervenes on that of their components: thus, they are not implicitly answering 'yes' to my question.) This second is the problem of downward causation.

    The problem, as I see it, is to come up with actual examples. As I said earlier, some philosophers of chemistry claim to have good reasons for saying yes, and some philosophers of physics claim the same in classical statistical mechanics. I don't understand these arguments -- it would be tremendously helpful if someone could explain them. However that might be, I am sceptical of the argument in biology, ecology, etc.

    So it's not a question of burden-shifting as I see it. I think most people acknowledge that no-emergence, no-downward-causation are the null hypotheses. But they think they have compelling reasons for rejecting these hypotheses. It's just that I have not come across a convincing argument in the fields that I just mentioned. The arguments are usually of the no-effective-way-of-restating-in-the-base-science variety.

  14. I agree with Matt that about the rhetoric here. Look, there are some prima facie obvious bridge principles that can be explicitly named. If X is a skeptic about them, (s)he needs a defense of this view.

    Something like the "more interesting" arguments by John Earman might do the trick though... Mohan, would you mind passing on that reference? I wasn't aware of an argument by Earman on this topic.

  15. Bryan, Sorry, I meant Larry Sklar, "The Reduction(?) of Thermodynamics to Statistical Mechanics," Philosophical Studies, xcv(1999):187-202.

    BTW, did you mean you agreed with me (rather than with Matt) about the rhetoric -- i.e. that since the no-emergence, no-downward-causation arguments are null hypotheses, burden-shifting isn't involved?

  16. I doubt I can give a knock-down style argument, but I can at least give you a thought on why I think biology cannot be reduced to physics:

    I hold that natural selection changes species over time and does so because it is, or applies, some sort of force. I have trouble seeing how this force can be reduced to any other.

    Anyone who doesn't see natural selection as such can readily disagree with me, but then has the task of explaining how/ why species change over time.

  17. Whoa Noah! What an example!

    Actually, I have written extensively on why natural selection is not a force, and in doing so have certainly tried to leave in place Darwin's theory of why species change over time. Very briefly, they change over time because individual organisms leave offspring in different numbers. The cause of their doing so is not some irreducible force known as natural selection, but just whatever causes individuals to leave descendants as they do. To wit: they are attracted to one another, they mate, they birth, they cope with their environments, they die, etc.

    There's a bunch of stuff about this on my web-page if you are interested:

    1. Mohan, you might be interested in my new book at You will quickly see I agree with Lakatos that selection is non-predictive (its just whatever may survive for whatever sensible reason).

      I look at the other side: at DNA using specific chemicals from the lanscape to build anatomies, and the limits to those capacities, for example flowing liquids (a heart), expanding gasses (lungs) used for an anatomy...for the very basics of chemical capacities.

      Thus predict from the capacities available for building, and then apply selection to weed out the duds.

  18. In all the special sciences it is acceptable to invoke entities and processes from more fundamental sciences in explanations. For example, the economy may be affected by the weather, living systems may be affected by radiation, chemical reactions may be affected by magnetic fields, and so on. There is a fundamental asymmetry between (fundamental) physics and the special sciences since it is never acceptable to invoke special science objects, events or processes to explain phenomena in fundamental physics. So in that sense at least some sciences are more basic than others.

  19. @James Ladyman - Nevertheless, you can't actually explain the movements of most physical objects in our vicinity without reference to fluid dynamics, chemistry, biology, psychology, economics, etc. (Neurath and Cartwright do an excellent job with generating examples of the right sort.) Physicists are of course dismissive of such examples, but dismissiveness doesn't help explain.

  20. Mohan, I think we probably agree much more than we disagree. But...

    My views are epistemically based: I believe that nothing within the theory of evolution or natural selection depends upon knowledge of physics. So we could forget all of physics and retain our knowledge of evolution.

    For example, we might lose genetics if we forgot our physics. However, evolutionary theory does not depend on genetics: they are evidence, not core propositions of the theory. Darwin came up with the struggle for survival and natural selection with no knowledge of the physical mechanisms, and they would remain the fundamental concepts of the theory regardless of what happens in physics. As such I see biology as epistemically independent of, and hence irreducible to, physics.

    The fact that you can give biological phenomena a physical explanation is a separate issue. If and when you do this I think you stop talking about biology and start talking about physics; not because you have reduced one to the other, but because you have just starting talking about something else. No one says that we have 'reduced' physics to biology when we described someone who has turned very pale as frightened, even though this is an evolutionary explanation describing a physical color change.

  21. Noah, you are right that we agree more than we disagree.

    I said in my original post: "The supervenience (or inclusion) claim may be of no interest to the chemist, since s/he has autonomous methods of investigating the domain. Further, it may be of no interest to the physicist, since it is messy and, moreover, reveals very little of interest concerning the physics-domain." And this seems to complement what you say about evolution: we can understand it without understanding physics because we have autonomous -- ie, physics-independent -- ways of investigating evolution.

    But as you acknowledge, this claim of yours is epistemic, and thus very different from saying that natural selection applies a force which cannot be reduced to physics.

    Matthew, I probably should let James Ladyman speak for himself, but "physical object" can be taken in two ways: as equivalent to (a) "material object"or (b) "physics-object". Physics only deals with atoms and their constituents, and you can't invoke chemistry or biology to explain the properties of these.

  22. Well two scenarios come to mind, one kind of reduction is a (true, strongly?) eliminative reduction, I show the morning star is the evening star and therefore eliminate one star from our ontology. Suppose, supervenience type reduction were like this, there was evidence for X, X would be eliminated by the putative reduction then you had better have evidence that the reduction works (complete bridge laws).

    However the supervenience relation does not work like that, it works more like an identity relation as far as I can see. Element valences (which determine common binding ratios for elements) are not shown to not exist by reduction to electron sharing patterns, rather they are elaborated upon. My only problem with saying that physics is more basic is that identity relations are usually two-way and I don't see why supervenience is not just an identity relation and that a statement about water is not as basic as a statement about H2O. However, physics is clearly more "complete" in that it takes all phenomenon as potentially representable/explainable under its system, also usually physics demands more accuracy (other theories may be deliberately vague/inaccurate).

    There is another kind of reduction, which I would call reduction of explanation. This is where you have some higher order theory that explains some general phenomenon (not specific instance). The claim will be made this explanatory work can be done by more basic physics. If perfect bridge laws (ie formal algorithms) existed to convert from higher to lower theory then this would be unproblematic and reduction would follow.

    However, in reality such formulaic bridge laws may well be impossible in all cases. Since for example Thermodynamics is actually just a different theory than Statistical Mechanics, they should not really be interconvertable. Presumably Thermodynamics is just wrong. However it may be wrong in a useful way. It may identify a class of phenomenon that do actually resemble each other in important ways and admit of a unifying explanation when described in those terms. There may be no formal move that will allow you to get a description from Thermodynamics back in rigorous statistical mechanics, and instead you will have to make some informal move (think Lakatos) that is equivalent to going back to Thermodynamics.

    There is no doubt that the more basic (or general) physics will instantiate all the cases, but it may not give you the unifying umbrella. This at least is my gloss of what Bob Batterman talks about.

    Allan Olley

  23. If you grant me my epistemic independence and you also hold that, "It may even be the case that there is no effective way of stating the supervenience relation," then I don't really see how there is much behind the claim that physics is ontologically basic. All it amounts to is a weak supervenience claim.

    Am I missing something?

  24. @Mohan - Let's take a prototypical object of classical physics: a canonball (the story works for a ball rolling down a ramp, or a pendulum, or electrons in a two-slit experiment, or a Bose-Einstein condensate, or whatever, mutatis mutandis.). Supposing that it isn't a windy day, etc., once the ball leaves the cannon at t=0 & x=0 with a certain momentum, you can predict for some range of subsequent times the trajectory of the ball using strictly physical properties and laws. BUT, for a range of times prior to t=0, you cannot explain the purely physical properties of the cannonball (position, momentum, etc.) without recourse to chemistry (combustion), material science, economics, political science, sociology, psychology, etc. When you say "you can't invoke chemistry or biology to explain the properties of" physical objects, if you mean, physicists aren't interested in explaining these kinds of situations, I agree! But there is no other way to explain them, and explanations that invoke the special sciences actually tend to work pretty well.

    So, Ladyman could be making the (probably) true but uninteresting sociological claim: physicists refuse to deal with circumstances that require appeal to the special sciences. Or, the apparently false claim: all physical objects (atoms, cannonballs, pendulums) and their properties can be explained without recourse to information from the special sciences, or that they cannot be explained with recourse to the special sciences. Or, the question-begging argument: special sciences can rely on explanations from the more fundamental sciences, but not vice versa, because those sciences are more fundamental. (Question-begging because their fundamentality is itself in question when we doubt their ontological priority.)

  25. The epistemology/metaphysics divide is clearly crucial here: even if not all scientific *explanations* can be given in purely physical terms, the *things* involved in the phenomena to be explained may well be nothing over and above aggregates of physical building blocks (be they simple and fundamental or not, the possibility of gunk, ontological infinitism etc. can be left open here). It was suggested in the thread that X's view may be: given naturalism, the epistemic lack of bridge principles justifies scepticism about the idea that physics is ontologically fundamental. But this seems wrong: a naturalist would be at least equally justified in taking the fact that physics is the science of the 'smallest constituents' as sufficient for claiming that the physical is ontologically basic.
    On the other hand, physics itself shows that it is not always the case that the whole supervenes on its smaller parts: I am thinking about the well-known example of quantum entanglement (where the state of the whole is simply NOT the result of the composition of the states of the parts). Couldn't X use this *in addition* to his/her claim about bridge principles to conclude that there is no reason for thinking that, as complexity increases, properties emerge that are not reducible to physics? Of course this wouldn't ground the claim that all sciences are equally basic in the sense that the domains of reality they explore are 'all ontologically on a par' (some basic properties would be existentially dependent on other basic properties but not vice versa), but it would in the sense that each science studies a domain of (at least partly) non-reducible systems and properties.

  26. Allan: supervenience isn't elimination. G is supervenient on F if two things must differ with respect to F in order to differ with respect to G. G.E. Moore introduced this notion in order to capture a certain kind of intuition about ethics: he said that goodness supervened on natural properties -- he certainly did not want to eliminate the good. On your other point, a suitable "nothing over and above" claim isn't nullified by the non-decidability of bridge laws.

    Noah: Of course, I grant "your" epistemic independence claim; after all, I stated it myself in the original post. And I also said that supervenience was not nullified by epistemological autonomy. There are lots of ways of talking about the basic ontological status of physics: material composition, supervenience, identity, etc. I didn't want to propose a new thesis about the relationship of physics to the special sciences: I expressly said that I was perplexed by the rejection of what seemed like a weak thesis.

    Matthew: I don't take a cannonball to be a prototypical object of classical physics -- point masses are, and they don't exist. Later on, atoms got drawn in. I don't take this to be a matter of sociology of science: it had more to do with the kinds of laws that constitute classical physics, the conditions of their applicability, and the methodological assumption of universality. You raise an interesting question though: what does classical physics apply to? My own feeling is that they apply to everything, but only under the point-mass idealization. By the way, the special sciences are explanatorily impotent with respect to point masses.

    Matteo: Right on! And as a footnote: I think that quantum entanglement might belong with what I have been calling "the more interesting arguments". These arguments point to specific failures of ontological dependencies. X's arguments dwelt on universal philosophical principles: no bridge laws, all representation is incomplete, and stuff like that. X might want to recruit these more interesting principles, but (I suspect) mostly on the "any port in a storm" principle.

  27. Ontological priority is not a question that deals between theses of different sciences, it is the method by which one determines what constitutes the being under consideration in any given case. e.g. Dawkins is an idiot because he doesn't recognize that during development of an organism the ontological priority changes from the genome to the cell structure, to the organism system as a whole. You are simply misusing the concept in discussing sciences.

    No science is "more basic" for a different reason. Each science deals with a different regional ontology to begin with. Each science has a posit (hence they are positive sciences) that declares the region of beings with which they deal. Physics is the science of motion, it is perfectly possible to do physics without the notions of matter and energy (information bits work perfectly well in many cases) as long as motion is posited. Biology is the science of life.

    Many of the difficulties the sciences have at the moment is that they have lost the basis of their posits. Biology keeps running into dead end ideas that result from mistaken assumptions regarding what constitutes life, and universities are unable to determine what ought to be required for a biology degree due to a lack of consensus on the nature of life and the appropriate methodologies required for its study. Much of the more recent work in the field focuses on emergent, non linear, self organizing systems, but I have yet to see a Biology program at a university include the requisite studies in non-linear mathematics and systems theory to understand this work.

    Psychology has no definition of the psyche to base a science on, since the definition originated in theology and is no longer accepted. The result is that new psychological 'theories' are not theories per se, but in reality attempts at founding new sciences, themselves lacking a basis in an accepted definition of the psyche. The result is that what is termed 'psychology' is nothing more than a collection of baseless pseudo sciences. In practice this eventuates the situation that there is no corrollory whatsoever between the psychological theory held by the practitioner and his/her success in practice, because the theories are baseless and irrelevant. Psychology as practice can be a beneficial or harmful social practice, depending on the practitioner, but it has nothing whatsoever to do with science.

  28. Well that's a conversation-stopper, Mitdasein -- so there's an end to it.

  29. Sandra Mitchell's recent book _Unsimple Truths_ argues (so far as I can see) exactly like X does in this thread (U of Chicago, 2009).

  30. Reading the above thread I feel we could perhaps avoid much confusion by distinguishing 'causation' (if such a thing exists) from 'models of causation' used to predict outcomes.
    In physics we have mathematical models of causation (e.g. f=m*a) which happen to predict outcomes with high accuracy. But prediction of outcomes with high accuracy does not necessarily mean that the model reflect the 'real' (ontologically basic) causal structure at work.

    For example, there is a theory, T, that 'Red lights cause cars to stop'. This predicts outcomes with high accuracy, but the predictive accuracy by no means confirms that the red lights actually cause the cars to stop. There may well be a more complex causal structure at work. The may even perhaps be no causal structure at work and we simply have regularity conjunctions (Hume's can't observe causation idea). Therefore the most we can say definitively is that theory T predicts outcomes with high accuracy. It is a model.

    If we agree about this then what is there to distinguish physics models from chemistry, biology, or sociology models? If we have no guaranteed access to the basic ontological reality underpinning observations then we cannot compare in terms of ontological priority.

    What we have are physics models that predict well for certain scenarios and then, when computations become too voluminous/complex, we start using chemistry or biology etc because we can't map out the maths for every single atom in these grand systems.

    All the arguments over how science 1 relates to science 2 seem (correct me if I'm wrong) to end up boiling down to an argument over causality. But all these arguments can be cleared up if we accept that causality (as per Hume) cannot be observed - it is only inferred and this inference is philosophically tenuous. Causal structures MAY exist (and I like to believe that they do) but we cannot access them. Without access we cannot compare science on the basis of what truly causes what.

    Would love to hear your responses.