Why biology (and chemistry) is not physics

In the Wall Street Journal, the physics writer Jeremy Bernstein has a fine review of a new joint biography by Gino Segre of George Gamow and Max Delbruck named "Ordinary Geniuses" which I just started reading.
Gamow and Delbruck were not as well known as some of their more famous peers but as Segre demonstrates, both made very important contributions to cosmology and molecular biology through direct experimentation and theorizing as well as by inspiring others' research. In addition the two were colorful and engaging characters which makes the book a pleasure to read.
But it was the first paragraph of the review that really caught my eye:

Some sciences are more unruly than others. Here's a parable to illustrate what I mean. Imagine that when the first life form appeared there was a superintelligent freak. If this freak had had a complete knowledge of the laws of physics, what could it have predicted? Quite a lot. All atomic nuclei consist of neutrons and protons, and the number of protons determines each element's chemical nature. Knowing this, the freak could have predicted all the elements that could possibly exist, along with their respective characteristics. Suppose that it also knew all the laws of biology, including the "central dogma," which explains how genes are expressed as proteins. Even so, it could not have predicted the existence of giraffes, nor even the fact that my brother and I share only half our genes. Both of these are evolutionary accidents. If it had not been for random mutation there would be no giraffes, and my brother and I might have shared all our genes, as male bumblebees do. Biology is not like physics.
This paragraph succinctly pretty much nails down the fundamental limitations of physics-based reductionism and it's a point that applies to chemistry as well. It's a very important point. The problem is that reductionism will never be able to account for the role of historical contingency and accident. Even if an all-powerful being could account for all biological scenarios emerging from an initial state of the universe, it could never tell us why one particular scenario is preferred over others. As Bernstein says, evolutionary accidents by definition cannot be predicted from starting conditions because they depend on chance and opportunity.
In addition function can never be uniquely derived from reductionism even if structure is. For instance in his book "Reinventing the Sacred", the complexity theorist Stuart Kauffman makes a powerful argument that even if one could derive the structure of the human heart from string theory in principle, string theory would never tell us that its most important function is to pump blood. The function of biological organs arose as an adaptive consequence of the countless unpredictable constraints that molded them during evolution. In addition the evolution of both structure and function was a mix-and-match process that depended as much on chance encounters as on strict adaptation. All this can never be captured in a reductionist worldview.
The same principle applies to chemistry. For instance the supreme being would never have been able to tell us why there are only twenty amino acids, why there are alpha amino acids instead of beta or gamma versions (which have extra carbon atoms), why amino acid stereochemistry is L while sugar stereochemistry is D, why there are four DNA bases with their unique structures, why nature chose phosphates (although Frank Westheimer comes close), why a given protein folds into only one unique functional structure, why water is the only solvent known to sustain life, and in general why the myriad small and large molecules of life are what they are. In retrospect of course one could provide several arguments for the existence of these molecules based on stability, function and structure but there is no way to predict these parameters prospectively.
The problem is that there is nothing in the nature of these molecules that dictates that their presence should have been uniquely determined. For instance we now know from synthetic studies that beta and gamma amino acids can also fold into the kind of helices and (less so) sheets that are ubiquitous for alpha amino acids. In addition these "higher order" amino acids provide extra handles for functional group attachment (see top figure). Yet for some reason they were discarded during evolution. Why? We could come up with several arguments. 

For instance because of their floppiness, maybe the higher order versions had to pay an unacceptable entropic penalty that could not compensate for their folding propensity. Or maybe the Strecker reaction that is thought to produce alpha amino acids could never be superseded by other chemical reactions for forming beta amino acids. Or perhaps alpha amino acids shield hydrophobic side chains much better than their longer chain counterparts. Cogent reasons, all of these, and yet I am sure we could find an equal number of arguments against alpha amino acids if we searched hard enough. The ultimate failure to find an explanation for the existence of alpha amino acids is a powerful reminder of the importance that chance and circumstance played in the evolution of both biomolecules as well as living organisms.
This role of contingency and accident is one of the most important reasons why the reduction of chemistry and biology to physics won't work. In addition as I have described before, reductionism cannot account for variety in chemistry. Yet another reason why chemistry and biology are not physics.


  1. Just a note: The "open dots" used as "attachment points" look an awful lot like O's. I had several seconds of confusion as to why a beta amino acid contained a peroxide ("well, *that's* not stable!")

    If you ever have an opportunity to redo the figure, I might recommend using filled dots instead.

  2. Done, thanks! (I wonder if a beta amino acid with the central carbons substituted by oxygens can be even fleetingly synthesized!)

  3. These types of questions/scenarios are especially important with origin-of-life science.

    I use the term "reductionist" differently from you, I think. I call using physics to predict biology "constructionism", while "reductionism" is looking at modern biology and figuring out what was there earlier in the process of evolution.

  4. Why can't Gamow and Delbruck's superintelligent freak being predict giraffes? If He/She knew all the laws of physics, why couldn't He/She have predicted the seemingly "random" events -- "chance" point mutations in proto-giraffe genes -- that led to the existence of giraffes? After all, those were simply caused by radiation damage to DNA, or mis-catalysis by a DNA-replicating enzyme, or ... -- in other words, something physical. It seems to me Gamow and Delbruck abandon the reductionist logic prematurely. Their assumption appears to be that truly random events do exist, but is that the case, or do they only appear random to use mere mortals?

  5. Yes, a superfreak could have predicted the set of all possible mutations. But there was still no way to decide which ones among those would prove beneficial and help the species evolve and propagate.

    However, your point about the perceived randomness of events is an interesting one. "Random" does not necessarily mean non-deterministic. In my head it has more to do with probabilities. Random events have probabilities that cannot be predetermined and therefore cannot be predicted.

  6. Given limitless computational power, it could have predicted giraffes as a possibility, among billions of other possibilities. It could not have said with certainty that giraffes, as we know them, would occur. Prediction through billions of bifurcation points is not possible. This is a key observation of chaos theory, in which small initial differences lead to widely divergent outcomes, rendering long-term prediction impossible.

  7. I have to say that I like Paul's usage of the phrase "constructionism." It suggests - at least to me - that while one should consider it necessary to be able to bridge physics to chemistry and chemistry to biology (and I suppose physics to biology as well), that it is not going to be sufficient to provide a complete understanding.

    Of course, I have to wonder if we're sometimes being overly demanding in expecting physics to lead into chemistry and/or biology - for example, the entire topic of protein folding seems to be a fairly popular one in the (bio)chemical blogosphere. There are classic physical systems that still invoke equally lively discussions in the literature (especially thinking of glass-forming systems here), despite having been available on the scientific research buffet for longer. And people are expecting equally or even more detailed physical pictures of protein folding?

    Having said that, I suspect it's a function of where one sits - I have the impression that the physicists (and physical chemists) who are interested in biological problems aren't going to be inclined towards explaining the chemistry of amino acids. They're going to be more interested in understanding in, say, signal transduction where they can control the strength of a signal (ligand concentration) and measure the output (some sort of enzymatic activity being up or down-regulated). If the receptor clusters, then they're off to comparing the Ising model vs MWC vs whatever else they can devise via simulations and subsequent comparison to experimental data.

    My two cents, change likely is warranted....

  8. In response to Wavefunction: Thanks for your reply. You say that "a superfreak could have predicted the set of all possible mutations. But there was still no way to decide which ones among those would prove beneficial and help the species evolve and propagate."

    But the theoretical superfreak can also predict the set of all possible mutations for *other* organisms in the system, not just the giraffe. Why can't He/She enumerate those possibilities simultaneously? This information would be the basis for decisions about which giraffe mutations would prove beneficial.

    This obviously implies an astronomical number of theoretical evolutionary pathways... but this is just a thought experiment anyway, so let's pretend He/She can evaluate each step of each pathway based on just physics. What information is lacking for this being to predict evolutionary history, unless we invoke truly random events?

    I don't quite follow your point on randomness vs. deterministic events with probabilities, so I'm not sure how that fits in... By the way, I don't want to undermine your article, because I think it's extremely interesting and provocative! I just wanna poke you about it a bit...

  9. FullyReduced: I appreciate your poking, it's a very interesting issue. The problem as I see it is that a lot of evolution has been governed by the propagation of events which might have appeared to be low probability events beforehand. Thus, even if the superfreak could calculate every single mutation in every organism along with every single environmental condition that could lead to these mutations being preserved, how could he/she/it know which one of those countless combinations will actually be the one that finally exists?

    You are right that among the countless scenarios predicted by the superfreak will be the universe and earth that we inhabit. But there is still no way to decide beforehand that this particular universe would be the one that actually materializes, part of the reason being that the a priori possibility of such a universe arising might be very low and there is no reason why the superfreak will pick a low probability event as the preferred one.

    MJ: I find your mention of other (poorly understood) classical systems interesting. As one example of why we may perhaps be overly demanding, consider that we cannot even accurately calculate the solvation energy for simple organic molecules (except for cases where you parametrize the system to death and use a test set that's very similar to the training set). With our knowledge at such a primitive level, it might indeed be overly demanding to try to predict protein folding which is orders of magnitude more complex.

  10. By the way, reductionism is supposed to imply the kind of constructionism (sometimes called "upward casusation") that Paul mentions. The fact that it does not speaks volumes.

  11. Interesting argument. I agree with you for the most part, though I feel that all of your chemistry examples are actually classified as "biochemistry."

  12. I suppose when one gets down to it, it's not just chemistry and biology - any system where one is looking at the behavior of many (interacting) entities is going to be complicated, and - I will be overly generous here - deriving its properties from first principles is going to be an extremely opaque process at best. People are still getting headaches from the entire "strong correlations in condensed matter" problem. Not being able to break out the "non-interacting, independent particles" approximation really irritates people. Especially when it fails to properly account for the properties that don't naturally fall out of said approximation. Heh.

    I do think, though, that the conceptual tools and formalisms that one develops in the physics can find fruitful new applications in biology and chemistry - although how much of that is just the unreasonable effectiveness of mathematics is always up for debate, I suppose.

    As a related followup to my previous comment in this thread, someone fortuitously sparked my memory today - there are those chemists incorporating parity violation into their calculations to explain chirality I remember hearing that the expected spectral differences might be too small to reasonably observe for the lighter elements spectroscopically, so they were starting to look at heavy element compounds.

  13. @Anonymous: Interesting point: "Prediction through billions of bifurcation points is not possible. This is a key observation of chaos theory..."

    But you use the term _prediction_, i.e. a guess from a human perspective with (by our very nature) limited data. In other words, it's not clear the findings of chaos theory negate determinism; rather, they seem to refute our ability to predict deterministic systems given our imperfect ability to gather information about the natural world.

    I guess what I'm trying to do here is separate out what's theoretically possible from what _we're_ capable of. (Of course, "theoretically" implies theories _we_ came up with, so maybe this is ultimately a dead end...) Thoughts?

  14. @WaveFunction: I think we may have different assumptions about what kind of predictive calculations this theoretical superfreak is capable of. You appear to assume mutations are truly random events and can't be predicted using the laws of physics. On the other hand, I assume mutations are determined by the physics of intermolecular interactions, incoming environmental radiation, etc. and that history proceeds in a stepwise fashion in which each step can be predicted from (1) the last step and (2) the laws of physics. (Of course, for this to be true, I'm assuming the superfreak has _perfect_ knowledge of the _true_ laws of physics, which of course we as a species do not currently have.) Thus the superfreak has knowledge of each mutation, and -- coupled with His/Her _complete_ knowledge of the environment -- can predict whether it will be retained.

    But that's a purely theoretical point, and I freely admit that large swaths of evolution were determined by (what _we_ see as) random events. From _our_ perspective, chaos theory comes into play here, as @Anonymous mentioned.

    Thanks again for the article -- good stuff here.

  15. Reduced: You are right that the findings of chaos theory don't preempt determinism; there is a reason why the field is termed "deterministic chaos". I have always found the line between the lack of prediction "in principle" and "in practice" somewhat fuzzy in case of chaotic systems. These systems are definitely (mostly) unpredictable in practice. But being predictable in principle would mean being able to specify the initial conditions of the system to an infinite degree of accuracy. I don't know if this is possible even in principle.

    MJ: Do you know of any parity-themed papers on chirality for the intelligent layman?

  16. @Wavefunction: Great point on infinite precision of initial condition defintions -- I feel that's finally the bridge between practical and theoretical limitations I was looking for.

  17. By the way: for posterity's sake, do you know what happened to my previous @Wavefunction post? It seems to be AWOL, and the conversation is kind of disjointed without it...

  18. The old adage: "the more you know about physics, the simpler it gets and the more you know about biology, the more complicated it becomes."

    However,fundamentally it is all physics. The fact that we cannot grasp the connection is our epistemological shortcoming. Moreover it is clear that Nature is not perfect so all it has to do is work. Maybe 25 amino acids will work better than 20, maybe a different protein fold along the way would not lead to cancer, but it does not matter. Eventually evolution will sort things out, given the right environment.

  19. I unfortunately don't know of any good review papers off the top of my head, but I would imagine if you search for Peter Schwedtfeger (the big-name theorist down in NZ), you'd eventually find something suitable. The entire "parity violation and chirality" topic was something that momentarily caught my eye when I was puzzling over a sideline topic a while back. From what I know, there hasn't yet been any experimental verification, although various metrology/precision spectroscopy groups are going after it.

    Also, something to think about in relation to chaos and being able to specify initial conditions - in classical systems, you describe your system in terms of its particles' position and momenta. Given that, one can specify said position and momenta exactly. When one moves into quantum mechanics, you suddenly now have a distribution in position & momenta that is a small "patch" in phase space that is proportional to Planck's constant, as one can only jointly localize position and momentum so far. I suppose this is why I find the mere notion of quantum chaos to give me headaches thinking about the evolution of little hyperblobs in six-dimensional phase space. Heh.

  20. I call false analogy.

    I think a better analogy for physics compared to the biology and chemistry examples given would be whether the supersmart freak could predict the number and location of stars in the galaxy/universe and how many planets are around each star.

    Chemists, knowing the fundamental laws of chemistry, can give knowledge of the properties and reactivity of as yet unknown compounds in much the same way physicists can with atoms.

    Biologists cannot be compared because biology deals with the specific system of life that has already arisen. To ask why this freak couldn't biology predict a giraffe when a giraffe isn't part of its biologic system is not an apt comparison in any way.

  21. Adding my voice to some of the others:

    If physics is deterministic (so let's suppose that a nonlocal hidden variable interpretation of quantum mechanics is correct), and if our superfreak (or Laplacian demon, as the more traditional account has it) knows the complete physical state in addition to all the laws (and has an unlimited computational capacity), then the superfreak will be able to predict the existence of giraffes and every other biological detail.

    Leaving out relevant physical details (i.e., the initial conditions) does not not show that biology is non-physical. Of course, it is true that physical dynamics alone will never tell us what sort of creatures evolve and which don't, but why would we ever think it might? Mere physical laws can't even tell us that there will be protons and neutrons.

  22. Physicalist, Bryan and Andre: Just want to make sure we carefully define what we mean by the reduction of biology to physics. It does not mean proving that all of biological matter is composed of basic subatomic constituents which is an obvious fact. It really means the "constructivism" that Paul was talking about. If we can truly reduce biology to physics, it must mean that we should be able to at least in principle do the opposite; construct the present biological world starting from the basic laws of physics.

    However it is not clear how we could go about doing this even in principle. The question is not just one of epistemology but of ontology. Again, I think Kauffman's example is very cogent. Even if the structure of the mammalian heart could be predicted in principle, it would be impossible to predict beforehand that the most important function of the heart among myriad others is to pump blood. The problem is not just epistemological in that we lack knowledge of all the conditions that could ultimately lead to a heart but ontological, namely that even if we had the knowledge we would be unable to assign probabilities to various scenarios. To me this seems to be the basic issue.

    I am not sure the existence of protons and neutrons was as subject to chance and circumstance as the evolution of the giraffe since it can be predicted based on very basic principles of energetic stability and knowledge of the five forces. So can the synthesis of the elements. But everything from then onwards seems much more subject to chance and accident.

  23. But you cannot necessarily construct the present physical world from the basic laws of physics, let alone the chemical or biological worlds. This is why the idea of predicting a giraffe doesn't seem to fit with the idea of predicting the elements.

    Could the planet earth be predicted from physical laws (not the life on earth, but the specific planetary make-up)? That is more akin to the giraffe example.

    More interesting (and IMO more appropriate) questions would be the following: Could, starting from the basic universal physical laws, complex or sentient life be predicted? Can the idea of biology itself be predicted?

  24. Question. Does the difference between biology (and chemistry) and physics boil down to the difference between inductive logic and deductive logic? Inductive logic (biology) reasons from a specific case to a general pattern, whereas deductive logic (physics) reasons from or applies general axioms and principles to a specific case. JR, Greenville, SC

  25. So did the laws of physics evolve or have they always "existed?"

  26. Anon 2: I am one of those people who think of a law as a compressed description of a set of regularities in nature. In this sense something that represents the law that we use must have existed since the beginning.

    Andre: I think that's a much better and more challenging question to answer and it takes us into all kinds of philosophical territory including the distinction between living and non-living. I am not sure physics could have predicted the existence of biology as we know it. But given enough time it probably could have led to aggregates of matter that demonstrate at least some features of life (at least growth). However from an ontological viewpoint I don't think physics could have predicted life since according to physics, life is nothing but a special but still uninteresting arrangement of quarks (or strings or whatever the physicists are calling it these days). There is no way a physicist could predict the various functions that the arrangement corresponding to, say a human being, could perform.

    Anon 1: To some extent yes and that has been the main problem with reductionism, although it has also been responsible for reductionism's phenomenal triumphs.

  27. Here is a longer response to anon 2 if it is not off topic. The answer to anon 2 is neither one of the alternatives. The laws of physics were created. WHEN THERE HAS NEVER EVER BEEN A CAR, this is like no one starting to build a car with no rules or ideas about what a car is, what it looks like, how it works, what it does, how it is different from a carrot, etc. On the other hand, if no one starts to build something, and something just appears by chance, the order and regularity that we see in the universe is astonishing. It is one thing if rules exist from the beginning for organizing creation according to statistics and chance. It is quite another if the rules themselves are the product of chance and appear out of "blind, thoughtless, mindless nothing”. How can a plan for the universe appear out of thoughtless mindless nothing? This is like waiting around for a rock that does not exist to have an idea. J. R. Greenville

  28. Interesting article, and very instructive as to the deep complexity of biochemistry. Thank you. But I do think the premise really just traffics in the different semantic usages of "deterministic," the level of certainty you ascribe to the laws of physics as currently understood, and the capabilities of your hypothetical observer.

    Let's say your "superfreak" has, in the religio-philosophical sense, complete omniscience but is still time-bound (i.e. does not simultaneously exist in the future as well as the past). If the "superfreak" has complete information about the laws of physics, AND has existed since the beginning of the universe, AND has the capacity to store and process information about every particle and wave function in the universe, then why couldn't he predict everything as it will actually turn out? The "random" mutation resulting from a "random" particle hitting a "random" atom in a "random" protein is only random if you assume that the "superfreak" hasn't followed each of those atoms, particles, etc. since the beginning of time.

    Of course, if one assumes that the "superfreak" is bound by laws of quantum mechanics as currently understodd -- so that uncertainy and probability are built in as part of the laws he "knows -- then it's true that he couldn't predict everything. But such a definition in the hypothetical makes physics, as well as biology, non-deterministic. At that point, all you're saying is that biology is "less deterministic" because it involves larger sets of particles, but each of those particles is itself fundamentally unpredictable outside of probability.

  29. Curious Wavefunction says:
    "If we can truly reduce biology to physics, it must mean that we should be able to . . . construct the present biological world starting from the basic laws of physics."

    Again, this form of "reduction" is just a non-starter. If you insist that we only have reduction when the laws (and only the laws) specify some feature of the world, then nothing can be reduced (except the laws themselves).

    The physical laws are compatible with the complete absence of matter, so the laws are never going to tell you whether there's matter or a total vacuum.

    The criterion for reduction that you're using is unhelpful, because on this account nothing can be reduced to physics.

    A much more useful account of reduction is one which asks whether we can predict some feature if we are given both the laws and the complete physical state (and unlimited computational power, since we're interested in ontology not epistemology).

    In this case, it seems clear that the Laplacian demon (superfreak) would predict the existence of giraffes (though the demon might not call them "giraffes").

    "Even if the structure of the mammalian heart could be predicted in principle, it would be impossible to predict beforehand that the most important function of the heart among myriad others is to pump blood."

    I know famous people like Fodor and Searle make this claim (I didn’t realize Kauffman did; I’ll have to look at his book at some point), but it’s just wrong:

    (a) Even if we insist that one needs to know the evolutionary history of a trait to know its “real function,” the Laplacian demon would have all of that information available. It knows the complete history of the total physical state of the universe.

    (b) If our demon (“superfreak”) is smart enough to care about which functions are “the most important” then it should have little difficulty recognizing that the function of the heart is to pump blood (even without peaking at the past). It would be able to recognize certain self-regulating processes that maintain themselves against the flows of entropy, and it would be able to recognize that the heart’s circulating blood is an important component of this self-sustaining process (whereas, for example, the sound the heart makes is not).

    Now, if we stipulate that our demon is not allowed to care about any structure or order above the level of particles, then you’re right that the demon will be ignorant of biological facts. But with this stipulation, the demon would also be ignorant of the shape of planets, the temperatures of stars, the rigidity of ice, and so on and so on. But this just shows that we shouldn’t make such a stipulation if we’re trying to figure out the ontology of the world.

  30. Re: most recent posts from @Anonymous and @Physicalist:

    I completely agree! I think I was stumbling to express similar thoughts earlier -- glad to see some backup. Thinking back now, the argument that initial conditions cannot in principle be precisely defined is really a limit to the superfreak's ability to predict *any* physical, chemical, or biological feature -- not just biological features like giraffes. So there's still no genuine distinction between the fields in that sense.

    Also, function is a much fuzzier concept that physical existence, which makes it hard to blame the superfreak for any inability to predict function ab initio.

    That said, to the extent that function can be defined, perhaps on the grounds of persistence of features/structures despite high entropy as Physicalist suggests (essentially a historical definition), the superfreak would have all the necessary information to make such a judgment, because He/She would know the entire physical history of the universe.

    Again, thanks for the fascinating thought experiment, Wavefunction, but I've ended up entirely unconvinced!

  31. I agree with FullyReduced's comment. Sadly the article is yet another example of a scientist misunderstanding the two different meanings of "reductionism". The first, simpler one, is theoretical and has to do with composition: if I take apart a person or a cell or even my alarm clock, I won't find any fundamental particles or forces unknown to physics. The second meaning is practical and has to do with explanation and prediction and suggests that all phenomena are best explained at the level of physics. The first is a bedrock of modern science and should be more widely promoted. The second is endorsed by essentially no scientists but is often confused by the public with the first. The first is the reason the superintelligent freak could (in principle) predict the entire history of life on earth. The second fails because we mere mortals can't. I wish scientists like this blog author and Kaufmann would stop doing a disservice to the public and be more clear about the two different meanings.

  32. I find this debate fascinating and want to thank everyone for contributing. Firstly with reference to FullyReducible’s point, I am pretty sure that I (and presumably Kauffman) are not confusing the two types of reductionism. In fact the first statement- that everything is ultimately composed of quarks of strings or whatever- is not even reductionism; it’s an obvious fact that nonetheless tells us nothing about complexity since there’s no context-specific dependence built into it For instance it cannot even tell us why two molecules with exactly the same atomic composition will have wildly different properties (again as a function of their environments).

    Now that we have gotten the first kind of non-reductionism out of the way, let’s focus on the second kind which matters. I don’t know why it’s so hard for the reductionists here to understand the difference between enumeration of all possibilities and the assignment of probabilities to each of these possibilities. I have already agreed that a superintelligent freak could list all of the countless events that would encompass the random mutations and effects of chance that we are talking about. But it would be impossible to assign a priori probabilities to all these events and predict that the net probability of our current universe existing is 1. This would be possible only if the superintelligent freak knows the entire future of the cosmos, in which case the discussion becomes meaningless and unscientific.

    Now let’s talk about function. I find FullyReduced’s statement about function being fuzzy very interesting (and it probably means we agree more than you think!) since that’s precisely why reductionism fails when it comes to function. It is precisely because ‘function’ is a result of the laws of physics compounded with chance that it’s difficult to predict on the basis of the laws alone. This leads into Physicalists’s objection that the Laplacian demon would be able to predict the function of the heart based on the environment in which it is embedded. But this environment itself is a result of countless chance events and encounters. So even if the demon could enumerate the many possible functions of the heart in advance, it would not be possible to say which function would turn out to be most important in our current environment. We are again facing the distinction between the a priori enumeration of possibilities and the assignment of weights to those probabilities. With reference to Physicalist’s last statement, it’s not so much that the demon is not allowed to care about structure, form and function but it’s really that she does not even know which form and function she should care about.

    This discussion also leads into Physicalist’s very interesting point about nothing possibly being reducible to physics since the laws of physics support a universe without matter. That is absolutely true. In fact that’s precisely why I find the idea of multiple universes, each compatible with the laws of physics, so alluring. Multiple universes will allow us to make a perfectly good case for non-reductionism without destroying the utility and value of the laws of physics. Extending the distinction between enumeration and valuation, it would mean that the laws of physics could indeed list every possible universe that can exist but are agnostic with reference to our own universe.

  33. This comment has been removed by the author.

  34. Read these comments and my thinking reminded me of this:


  35. @ Wavefunction:

    I wrote an (overly long) reply, but I couldn't get it to post here, even when I tried breaking it up into smaller bits.

    So I posted it over at my place.

    Short version: No scientific law deals with actual historical events; laws are about regularities. So your complaint has nothing in particular to do with physics.

    Once we include actual historical events (i.e., the physical state) with the laws, then physics has no trouble accounting for everything in the actual world.

  36. That's weird. Long comments do seem to have trouble getting posted. I will try to post your comment in bits and respond to it shortly.

  37. (Trying again:)

    @ Wavefunction

    OK, we agree that one form of reduction fails: It is not the case that all biological facts can be reduced to physical laws. Indeed, almost nothing can be reduced to the physical laws alone.

    Do we also agree that biological facts are reducible to a much broader category of physical facts? That is, if in addition to the laws we also include all of the actual physical properties (the complete physical state at all times), then all the facts about biology will be fixed. So once you know everything about the particles (so to speak), then you can – in principle at least – deduce everything there is to know about giraffes, hearts, and so on.

    It seems that at least some of our disagreement is over what it means to “reduce” something to physics. I don’t know whether it’s helpful to point this out, but scientists and philosophers typically mean different things by “reductionist.” Scientists typically call something a “reductionist account” if the relevant mechanism lies at some lower level (molecular biology, say). If instead the relevant mechanism involves whole populations or ecosystems, then the account is said to be “anti-reductionist” or “holistic.”

    Philosophers, on the other hand, are more interested in whether the higher-level facts can be derived from the lower-level facts (or whether there are higher-level causes that are independent of physical causes). Most scientists think that the philosopher’s reduction is an obvious truth and so isn’t even worth considering (it just rules out magic and ghosts, they’re inclined to say), and so they instead focus on questions of complexity, emergent regularities, etc. Which is all well and good, as long as one isn’t thereby seduced into thinking that structure, function, biology, etc. somehow go beyond the purely physical (which is the philosopher’s notion of reduction).

    There’s also a lot that could be said about different philosophical varieties of reduction. Nagel, for example, was mostly interested in the reduction of one theory to another. This would tie in with your attention to laws (and seeming indifference to states/properties). But the examples you embrace (e.g., the existence of giraffes) don’t seem to fit in too well here, since they seem like specific happenings rather than general laws.

    In any case, I’m a hard-core reductionistic physicalist, but even I don’t buy into the claim that you can derive one theory merely from the laws of another theory (and some “bridge principles”). It seems obvious that you’ll also need to know some details of what actually happened – that is, you’ll need to know some physical details in addition to the laws.

    A superintelligent freak could list all of the countless events that would encompass the random mutations and effects of chance that we are talking about. But it would be impossible to assign a priori probabilities to all these events and predict that the net probability of our current universe existing is 1. This would be possible only if the superintelligent freak knows the entire future of the cosmos, in which case the discussion becomes meaningless and unscientific.

    If physics is deterministic, then knowing the complete physical state and the physical laws does indeed provide knowledge of the complete future of the cosmos (setting aside epistemological issues of computability, possibility of measurements, etc.). I don’t see how this in any way makes the discussion “meaningless” or “unscientific.”

  38. (continuing: )

    Now, if you’re not interested in ontology, but instead in what we very finite and flawed creatures could come to know, then of course our knowledge of physics is going to tell us little about biology. I think we all agree on this. But this is due to our ignorance of some relevant physical facts (e.g., how some DNA will – as a natter if fact – recombine during a particular case of meiosis); it is not due to some biological fact that goes beyond the physical facts.

    Of course, if physics is indeterministic, then knowing the complete state at a time and all the laws still won’t allow one to precisely predict the future. But even in this case, there’s an important physicalist principle that holds: One will never be able to make a more accurate prediction using a higher-level science (e.g., biology) than one would in principle be able to make using the complete (indeterministic) physical account. Any higher level account is going to leave out some physical details, and therefore will be less accurate (though perhaps more illuminating) than the physical account.

    This is why I find it very misleading to claim that a reduction to physics fails because the existence of some things depends on chance events. To the extent that the claim is correct, it is completely irrelevant to understanding the relationship between physics and higher-level sciences like biology.

    The key point is that scientific laws are generally not in the business of telling us what exists. Laws (for the most part) tell us “If you start in state A, then you will end up in state B.” This is true of physics, but it’s also true of all other scientific laws. Laws don’t tell you what exists; they just tell you how one state time-evolves into another.

    So if physics has a problem with chance events and historical contingency, than so does every other scientific law – but really it’s not even about “chance” or “contingency” at all, it’s just a simple fact that the laws of nature are (as best we can tell) compatible with many different histories of the universe. There’s more to the world than just the laws. (If the world is deterministic, then there’s also the initial conditions. If the world is indeterministic, then we also have the outcomes of all stochastic processes.)

    So if by “physics” we mean to include physical descriptions of actual physical stuff around us (let’s just pick some hydrogen atom in space), then physics will indeed include an account of “historical contingency.” Even in a deterministic world, that hydrogen atom is there because of the history of the universe. In an indeterministic world, there will also be chance physical events. So a complete physical description will also take “accident” into account in that it will include the outcomes of stochastic processes.

    But once we recognize this, then there seems to be no reason to single out particular biological features (such as the function of the heart or the existence of giraffes) as being “not reducible to physics” just because they appeal to some historical fact in addition to the laws of physics. This doesn’t distinguish functions, or biological features, or complexity from anything else. All we’re saying is that the laws don’t tell us what the actual physical state is.

    Thus pointing out that the “environment itself is a result of countless chance events and encounters,” seems to be beside the point. If the super intelligent freak (or Laplacian demon) knows all the physical facts (including the outcome of any indeterministic physical processes), then the freaky demon will know everything there is to know about the environment, and the organism, and the population the organism belongs, to and the complete history of that population, and so on.

  39. (and the third and final part: )

    This is why I’m suspicious of your claim that “a superintelligent freak could list all of the countless events that would encompass the random mutations and effects of chance that we are talking about. But it would be impossible to assign a priori probabilities to all these events and predict that the net probability of our current universe existing is 1.

    Of course the freak would be unable to assign any useful probabilities without knowing the actual state of the universe. But why should we suppose that physics is about “a priori probabilities” (to use your phrase)? Physics is about transitions from one state to another state. Neither or knowledge of the laws nor our knowledge of the states is a priori. And, of course, no one would expect that you could derive even the probabilities of a final state without knowing something about the initial state.

    Further, I see no reason to suppose that a reduction of biological facts to physical facts would require determinism (i.e., the claim that given the actual initial state, the probability of landing in the actual current state is 1). Let’s allow for an indeterministic quantum physics. Then the complete physical story is going to include the outcoc

    The key point as far as physicalism goes is that given the outcome of indeterministic physical processes, there’s no further leeway regarding biological facts. It’s never going to be the case that a freaky demon knowing all the biological facts will be able to predict some future event more accurately than will the freaky demon who knows all the physical facts (both the laws and the complete physical state). Thus physics is the most complete, most accurate, account of our world, and there’s no extra wiggle room left once the physical facts are given.

    I haven’t read Kauffman, but usually the complexity folks are particularly interested in how the behavior of some systems is largely (though obviously not completely) independent of initial conditions, whereas the behavior of other systems depends sensitively on the precise details of the initial state (e.g., chaotic systems). There are many interesting issues to pursue here, but think we only obscure them if we suggest that they somehow take us beyond physics.

  40. (I think some unclosed html tags might have been the main problem -- in addition to the length restriction. The first third did show up when I posted it, but now it seems to have disappeared. I'll not try to repost it since it might make even more of a mess of things . . .)

  41. Thanks for your detailed reply. I finally got around to reading it. Basically the thrust of your argument and that of some others here seems to be that it's unfair to blame physics (or reductionism) for being unable to predict chance events dependent on historical contingency.

    But if this is true (and I agree that a failure to predict one specific case does not automatically imply a failure to predict the general case), then doesn't it pretty much void the entire utility of reductionism? Since pretty much every fact about the present universe (and certainly most of biology) has somehow been influenced in a major way by chance, a failure of reductionism to predict these facts would render the doctrine almost impotent. The only kind of "reductionism" we will be left with then is the hollow and obvious declaration that everything is made up of quarks or strings and that these "somehow" must account for everything that we see in our universe.

    The second point concerns your assertion that the kind of reductionism you are talking about will be able to actually give an account of the prehistory of a particular hydrogen atom somewhere in the universe. But stating the prehistory is far from being able to state the reason for its existence. In this context I can do no better than David Deutsch who interrogates exactly this claim of reductionism in his "The Fabric of Reality". Deutsch talks about a particular atom of copper residing at the tip of a particular statue of Churchill, and I think he really nails it when he says the following (sorry for copying out his words at length but I think they really capture the problem):

  42. “For example, consider one particular copper atom at the tip of the nose of the statue of Sir Winston Churchill that stands in Parliament Square in London. Let me try to explain why that copper atom is there. It is because Churchill served as prime minister in the House of Commons nearby; and because his ideas and leadership contributed to the Allied victory in the Second World War; and because it is customary to honor such people by putting up statues of them; and because bronze, a traditional material for such statues, contains copper, an so on. Thus we explain a low-level physical observation-- the presence of a copper atom at a particular location-- through extremely high-level theories about emergent phenomena such as ideas, leadership, war and tradition.

    There is no reason why there should exist, even in principle, any lower-level explanation of the presence of that copper atom than the one I have just given. Presumably a
    reductive 'theory of everything' would in principle make a low-level prediction of the probability that such a statue will exist, given the condition of (say) the solar system at some earlier date. It would also in principle describe how the statue probably got there. But such descriptions and predictions (wildly infeasible, of course) would explain nothing. They would merely describe the trajectory that each copper atom followed from the copper mine, through the smelter and the sculptor's studio, and so on. They could also state how those trajectories were influenced by forces exerted on surrounding atoms, such as those compromising the miners' and the sculptor's bodies, and so predict the existence and shape of the statue. In fact such a prediction would have to refer to atoms all over the planet, engaged in the complex motion we call the Second World War, among other things. But even if you had the superhuman capacity to follow such lengthy predictions of the copper atom's being there, you would still not be able to say, 'Ah yes, now I understand why it is there.' You would merely know that its arrival there in that way was inevitable (or likely, or whatever), given all the atoms' initial configurations and the laws of physics. If you wanted to understand why, you would still have no option but to take a further step. You would have to inquire into what it is about that configuration of atoms, and those trajectories, that gave them the propensity to deposit a copper atom at this location. Pursuing this inquiry would be a creative task, as discovering new explanations always is. You would have to discover that certain atomic configurations support emergent phenomena such as leadership and war, which are related to one another by high-level explanatory theories. Only when you knew those theories could you understand fully (emphasis mine) why that copper atom is where it is.”

  43. I would be very interested in knowing what you think about Deutsch's argument since it leads to your third very interesting objection, namely that even if lower-level phenomena may not be entirely accurate in explaining a particular fact, there's no reason that higher-level phenomena should be any more accurate.

    I do think that there are several cases in which higher-level phenomena are more accurate in capturing the essence of a fact. Consider for example the question of why DNA (and not some other molecular entity) is the genetic material. Ignoring for the moment that this question may never be answered accurately because of the role of historical contingency in the evolution of DNA, it's pretty clear to me that an explanation based on the chemical properties of DNA (hydrogen bonds, specific base pair interactions etc.) is not just far more operationally useful but also far more relevant. But it's not just this. One will have to consider even higher-level disciplines like population genetics and ecology to understand why DNA is the preferred genetic material. In this case it's pretty clear to me that higher-level phenomena are far more accurate in answering the "why" of a particular situation than lower-level phenomena. So while I do agree that higher-level phenomena are not always more accurate than lower-level ones in explaining a particular fact, I think there are several cases in which they succeed in giving us a more accurate description.

  44. Thanks for your response; I think we probably agree more than not. (But there still seems to be some "not.")

    I've posted up a (rather lengthy) response over chez moi, since it was quicker and easier than posting here.


  45. One of my favorite quotes in the Black Swan by Taleb, about predicting the motion of billiard balls.
    "If you know a set of basic parameters concerning the ball at rest, can compute the resistance of the table, and can gauge the strength of the impact, then it is rather easy to predict what would happen at the first hit. the second impact becomes more complicated, but possible. You need to be more careful about your knowledge of the inertia states and more precision is called for. The problem is that to correctly compute the 9th impact you need to take into account the gravitational pull of someone standing next to the table. and to compute the 56th impact every single elementary particle of the universe needs to be present in your assumption. Now consider thh additional burden of having to incorporate predictions about where these variables will be in the future."

  46. stumbled on this post in a google search for why chemistry/biology are not sciences, which they are not. i will help you guys out on the demon/freak. to quote a friend of mine, the problem is choice. now here's a real mind bender for ya: what if we can choose to change the laws of physics ;)

  47. Ash,

    I know this was an older post, but I somehow stumbled on it and really enjoyed reading it. I have to say, though, I disagree with your implicit assumptions about the randomness of environmental effects/mutation events, etc at a really fundamental level.

    The way I use the word "reductionism" is fairly straight forward. The universe operates on a set of input-output rules. That's it. If that's the case, then the proposed super-calculator could achieve a perfect prediction of every species to have existed, because each output state of the universe is a product of the prior state (the input).

    The only kink comes in because either we (finite computers) can't assess all the input variables, or don't know what they are, or because you are proposing random (non-functional-product) input. So which is it?


Markup Key:
- <b>bold</b> = bold
- <i>italic</i> = italic
- <a href="http://www.fieldofscience.com/">FoS</a> = FoS