Thursday, August 28, 2008

Reductionism

Richard Feynman once said that the two cultures of the sciences and the humanities, as famously described by C. P. Snow, are divided between those who understand mathematics well enough to appreciate its relationship to nature and those who have never had that experience. Although I don’t like the idea of there being two separate cultures in academia (however accurately it may describe the modern university), Feynman’s point is hard to deny. At a certain level, there can be no metaphorical understanding of the deep relationship between mathematics and nature, no way to translate the language of mathematics in order to convey its astounding predictive power in the physical universe. This must be experienced first hand, and the experience can change the way you see the world in profound ways.


I had a similar opinion about reductionism in biology. Reductionism is the idea that all levels of investigation in biology, from cells to organisms to populations and ecosystems, can ultimately be explained by molecular biology. This does not necessarily commit reductionism to what is commonly understood as genetic determinism, but it does imply molecular determinism. A reductionist would say that any meaningful explanation in biology will ultimately come from a description of how molecules at the biochemical level determine the phenomena in question. Non-reductionists—surprise, surprise—deny that all explanations in biology can be made in terms of molecules. They claim that some properties of biological systems “emerge” and that we cannot explain this emergence by a reduction to chemistry and physics.


Like Feynman, I suspected that the difference between reductionists and non-reductionists was a difference of experience. What separated the two, I thought, must be the level of familiarity with molecular and developmental biology. Those who had a cursory understanding of the relevant biology could be forgiven if they thought genetics and biochemistry couldn’t explain embryology, couldn’t explain adaptation, couldn’t even explain Mendelian genetics. Those who knew the facts would be convinced that reductionism was hard to deny.


Apparently, my suspicion was wrong. Or, rather, I was misled by my unintentionally selective reading of prominent biologists and philosophers. The issue of reductionism per se does not come up in biology textbooks, so my understanding of the larger conceptual framework of modern biology came from reading the few popular reductionists of our day: Watson and Crick, E. O. Wilson, Dawkins, and Dennett. I had been reading these authors before I took genetics, biochemistry, and developmental biology as an undergraduate, so it seemed natural to me that reductionism was alive and well in the realms of theoretical biology and the philosophy of science.


According to Alex Rosenberg (Darwinian Reductionism:Or, How to Stop Worrying and Love Molecular Biology), there are very few biologists and philosophers who are openly reductionistic. He finds this strange, considering the fact that the vast majority of these same non-reductionists are physicalists. The term physicalism is in conceptual opposition to the term dualism, an idea which Descartes is famous for. Dualism was an attempt to split reality into two different kinds of fundamental stuff—the physical world and the world of the mind or spirit. This idea encouraged progress in the physical sciences because it allowed philosophers/scientists to treat the physical world as a purely mechanical system which was accurately described by mathematics. Dualism, however, has fallen out of favor in philosophy and especially in the natural sciences. The main problem with dualism turned out to be the need to connect the two kinds of stuff. It seemed obvious that a person’s mind acted on its body (I mentally choose to lift my arm) and the body acted on the mind (I stub my toe and I am acutely aware of the pain in my mind) but no mechanism was discovered that could explain how this worked. Physicalism solves this problem by claiming that there is only one kind of stuff, the physical, and that the mind will eventually be explained by matter and its motion in space—a claim that has plenty of problems of its own.


What Rosenberg finds strange is that the majority of biologists and philosophers maintain the physicalist position and at the same time deny reductionism in biology. Again, physicalism is a mechanistic theory, or, I should say, it is a theory that there is one kind of stuff whose actions and reactions can be consistently described. This was done very successfully with Newtonian mechanics until the early 20th century. Today, physics is even more precise in its description of matter in motion, but it no longer describes a clockwork universe. Subatomic physics can only be described statistically, and this changed our conception of a purely physical universe which operates by consistent mechanical laws concerning matter in motion. However, quantum mechanics is still a physicalist approach to describing the one kind of stuff in our universe. Our understanding of quarks and electrons determines our understanding of atoms which determines our understanding of molecules and their chemical reactions. There is no gap in the line of explanation, which rests on our understanding of the brute statistical facts of subatomic particles: molecules and chemistry reduce to quarks and leptons. Non-reductionist biologists claim that there are facts in the biological world, from the level of the cell to the ecosystem, that are not explained by a line of reasoning that eventually leads to the quark and leptons. They claim that there are facts about the physical universe that are biological facts and essentially distinct from molecular/chemical facts. We are overdue for an example.

Meiosis is the process by which a cell splits to form gametes, which are units of sexual reproduction (sperm or egg).



http://en.wikipedia.org/wiki/Image:MajorEventsInMeiosis.jpg

The objects inside the cell in the picture above are chromosomes containing DNA, and the process of meiosis is responsible for copying the entire genome of an organism so that a complete set of chromosomes is placed in each gamete. This process is fairly well understood at the molecular level, and it has been used for decades now to explain Mendelian inheritance. Mendel was the Russian monk in the 19th century who cultivated thousands of pea plants, and from his observations developed a theory of inheritance that postulated units of inheritance called alleles. These alleles were thought to be responsible for traits like color, size, and texture. His theory stated that each allele had two versions, one dominant and one recessive, and that each plant possessed two copies of each allele. Depending on the combination of alleles possessed by each pea, two peas are bread and produce a certain ratio of offspring that express the traits according to the presence of dominant alleles.




http://ib.berkeley.edu/courses/ib162/Week2a.htm

The picture above illustrates the inheritance of color in a pea. The two peas outside of the grid are the parents, and the arrows indicate the contribution of alleles from each pea. Each parent pea possesses a dominant allele (Y) and a recessive allele (y). In this case the dominant trait is the color yellow, hence the capital Y, and the presence of at least one dominant allele will give the daughter pea a yellow color. In the absence of a dominant allele, a green pea will be produced. Since each pea contributes one allele to a daughter, and there is a 50/50 chance that the allele contributed will be a Y or a y, then the grid above represents all of the possible combinations that could arise along with their frequencies. In other words, if there are 100 offspring produced from these two parents, there will likely be 25 YY offspring, 50 Yy, and 25 yy. This is a statistical result, so the numbers won’t be perfect, but the likely number of yellow offspring peas will be 75. This concept of parents having two alleles for each trait, which are randomly segregated and distributed to offspring, is known as the Law of Segregation.


The molecular mechanisms of meiosis seem to explain Mendelian genetics by identifying alleles as genetic units. In the picture of meiosis above, the chromosomes represent many, many alleles, which are matched by a “homologous” chromosome containing the other allele. For example, the human cell contains 46 chromosomes.




Each chromosome, containing a series of alleles, can be paired up with one other chromosome that contains a complementary series of alleles, as displayed in the picture above (22 chromosomes are matched by size and color—the 23rd set contains the sex chromosomes X and Y). For simplicity’s sake, imagine that chromosome 1 contains an allele for hair color, with (B) as the dominant allele yielding brown hair and (b) as the recessive allele. If hair color in humans was produced in children at the same frequencies of color found in Mendel’s peas, then it would make sense to predict that two parents with both a dominant allele and a recessive allele (Bb) x (Bb) will most likely have offspring of the frequency displayed in Mendel’s peas: 75% will have blond hair and 25% will have no B allele, yielding, let’s say, red hair. Most traits in humans don’t have such a simple outcome (including hair color), but that’s the basic idea behind inheritance in peas—alleles are genetic units located on chromosomes.


Now let’s consider two traits that are simultaneously passed on from parents to offspring. Imagine that the same two parents with both dominant and recessive alleles for blond hair on chromosome 1 (Bb) also possess both dominant and recessive alleles for big ears on chromosome 2 (Ee). The cross would be written as (Bb) (Ee) x (Bb) (Ee) and there would be four possible types of offspring: blonds with big ears, blonds with small ears, red-heads with big ears, and red-heads with small ears. What should we expect their frequencies to be?


When Mendel tested for two traits, color and roundness, he found that, if he bred peas that he suspected contained both dominant and recessive alleles for both traits [roundness, R, is a dominant trait—so by crossing (Yy) (Rr) x (Yy) (Rr) he had to have a reason to believe they contained those alleles, since the outward appearance of the peas would all be yellow and round, the same appearance of a pea with alleles (YY) (RR), etc.] he obtained a result that was consistent with the assumption that the alleles were not connected to each other. In other words, when considering all of the possibilities of offspring, it was accurate to assume that the alleles for color were distributed independently of the alleles for roundness.




http://ib.berkeley.edu/courses/ib162/Week2a.htm

The four types of offspring were yellow-round, green-round, yellow-wrinkled, and green wrinkled, at a ratio of 9 : 3 : 3 : 1, but each trait could be considered independently as before. Seventy-five percent were yellow, 25% were green, 75 % were round, and 25% were wrinkled. This is known as the Law of Independent Assortment.


Going back to our human example, it is clear why we would expect the same frequency as Mendel’s peas. We should get blond-big-ears, redhead-big-ears, blond-small-ears, and red-head-small-ears at 9 : 3 : 3 : 1; and 75% blond, 75% big-eared, 25% red-headed, and 25% small-eared. Why? Because hair color alleles and ear size alleles are on separate chromosomes.


When meiosis occurs, each homologous chromosome is randomly split among the gametes that are produced, making the likelihood of getting blond or red hair and getting big or small ears completely separate. If they were on the same chromosome, then we would not observe the same frequency as before. The likelihood of getting blond hair would be statistically connected to the likelihood of getting big ears. This molecular explanation seems to account for Mendel’s non-molecular observations more than a century ago: all independently assorting traits are on separate chromosomes [or they are far enough apart on the same chromosome for cross-over to cancel out the affect on the frequency].

This is all necessary to understand because the arguments against reductionism have been lodged at the very first step in biology, the step from biochemistry to cytology, which is the study of the cell. Technically, the process of meiosis is a cytological process, because it looks at the patterns of molecular behavior that comprise a certain function of the cell. In this case, the function is replication for sexual reproduction. Philip Kitcher (“1953 and All That: A Tale of Two Sciences” and “The Hegemony of Molecular Biology”) has argued that classical (Mendelian) genetics, as a cytological process, cannot be reduced to molecular biology. His beef is not with what I have been describing as the molecular explanation of chromosomes as the location of alleles. He is denying the possibility of explaining the entire process in terms of the chemistry of molecules. ‘Simply’ describing the atomic makeup of molecules and understanding the chemical details of their reactions will not, according to Kitcher, provide us with the kind of explanation that is required to understand classical genetics.

“We begin by asking why genes on nonhomologous chromosomes assort independently. The simple cytological story…answers the question. That story generates further questions. For example, we might inquire why nonhomologous chromosomes are distributed independently at meiosis. To answer this question we would describe the formation of the spindle and the migration of chromosomes to the poles of the spindle just before meiotic division. Once again, the narrative would generate yet further questions. Why do the chromosomes ‘condense’ at prophase? How is the spindle formed? Perhaps in answering these questions we would begin to introduce the chemical details of the process. Yet simply plugging a molecular account into the narratives offered at the previous stages would decrease the explanatory power of those narratives. What is relevant to answering our original question is the fact that nonhomologous chromosomes assort independently. What is relevant to the issue of why nonhomologous chromosomes assort independently is the fact that the chromosomes are not selectively oriented toward the poles of the spindle. In neither case are the molecular details relevant. Indeed, adding those details would only disguise the relevant factor.”

Kitcher realizes that there is an obvious reductionist response to this. The problem may simply be that we do not know enough of the molecular details at the moment, but, if and when we do, it will eventually be possible to reduce meiosis to the biochemistry of cellular molecules. Kitcher thinks this response misses the crucial point:

“Recall the original cytological explanation. It accounted for the transmission of genes by identifying meiosis as a process of a particular kind: a process in which paired entities (in this case, homologous chromosomes) are separated by force so that one member of each pair is assigned to a descendant entity (in this case, a gamete). Let us call processes of this kind PS-processes…PS-processes are heterogeneous from the molecular point of view. There are no constraints on the molecular structures of the entities that are paired or on the ways in which the fundamental forces combine to pair them and to separate them. The bonds can be forged and broken in innumerable ways: all that matters is that there be bonds that initially pair the entities in question and that are subsequently (somehow) broken. In some cases, bonds may be formed directly between constituent molecules of the entities in question; in others, hordes of accessory molecules may be involved. In some cases, the separation may occur because of the action of electromagnetic forces or even of nuclear forces; but it is easy to think of examples in which the separation is effected by the action of gravity. I claim, therefore, that PS-processes are realized in a motley of molecular ways.”

This idea, that a certain kind of biological process can be accomplished using a variety of constituents and a variety pathways, is called multiple realizability. In other words, in the biological world, it’s possible to imagine that the process of meiosis could be successfully accomplished using different molecules and different chemical or physical processes. Kitcher’s point is that, since there is not a universal molecular description of the process of meiosis in all organisms, then it is impossible to think that the general process, meiosis, can be reduced to specific molecular processes. This is sometimes referred to as the many-one problem: many molecular processes can accomplish a specific higher cellular process.


Not only does Kitcher argue that higher levels of investigation, such as cytology and physiology, have “autonomous levels of biological explanation” and cannot be reduced to molecular biology, but he also rejects the idea that any exchange between the levels of investigation must flow from the bottom up. Kitcher claims that explanation can proceed from the top down.

“For example, to understand the phenotype associated with a mutant limb-bud allele, one may begin by tracing the tissue geometry to an underlying molecular structure. The molecular constitution of the mutant allele gives rise to a nonfunctional protein, causing some abnormality in the internal structures of cells. The abnormality is reflected in peculiarities of cell shape, which, in turn, affects the spatial relations among the cells of the embryo. So far we have the unidirectional flow of explanation which the reductionist envisages. However, the subsequent course of explanation is different. Because of the abnormal tissue geometry, cells that are normally in contact fail to touch; because they do not touch, certain important molecules, which activate some batteries of genes, do not reach crucial cells; because the genes in question are not ‘switched on’ a needed morphogen is not produced; the result is an abnormal morphology in the limb.


“Reductionists may point out, quite correctly, that there is some very complex molecular description of the entire situation. The tissue geometry is, after all, a configuration of molecules. But this point is no more relevant than the comparable claim about the process of meiotic division in which alleles are distributed to gametes. Certain genes are not expressed because of the geometrical structure of the cells in the tissue: the pertinent cells are too far apart. However this is realized at the molecular level, our explanation must bring out the salient fact that it is the presence of a gap between cells that are normally adjacent that explains the non-expression of the genes. As in the example of allele transmission at meiosis, we lose sight of the important connections by attempting to treat the situation from a molecular point of view. As before, the point can be sharpened by considering situations in which radically different molecular configurations realize the crucial feature of the tissue geometry: situations in which heterogeneous molecular structures realize the breakdown of communication between the cells.”


Kitcher’s arguments make sense, and they have apparently been influential among philosophers of science. However, I’m not convinced and, since I’m in danger of making this essay way too long, I’ll give the short version why. The fact is, the capacity for computation is the only limiting factor in our understanding of molecular biology, and we are not close to reaching our limit yet. It is entirely possible that we will be able to obtain an accurate and precise description of molecules in motion within the cell so that the functions now understood at the cellular level (cytology) will be reduced to biochemistry. Kitcher’s response to this doesn’t account for the possibility of an extremely advanced understanding of biochemistry that will in fact provide an explanation of meiosis. This seems like an anti-climactic conclusion after such a long introduction to this issue (I’m basically saying, “Nuh-uh”), but I’m frankly dumb-founded to find out that so many biologists and philosophers espouse the antireductionist argument. If the history of western science teaches us anything, it is that those who wager against the progress of reductionism in scientific explanation usually end up losing the bet. There are famous exceptions to that right now (for example, the mind), but I can’t see a good reason to doubt that molecular biology will not continue its “hegemony” in the life sciences.

Two final points. The first is that Alex Rosenberg has proposed a very interesting theory that may strengthen the reductionist position. There are many issues in this topic of reductionism that I haven’t even mentioned (for a comprehensive summary, see the online Stanford Encyclopedia of Philosophy entry on Reductionism in Biology), one of them being the role of natural law in scientific explanation. There don’t appear to be laws in biology like there are in physics, and this poses problems for explanation in biology and, in turn, reductionism. The closest thing to a law in biology is the principle of natural selection (PNS), and this theory hasn’t been successfully reduced to a physical law. Rosenberg suggests that the PNS is actually a law of chemistry and can be observed at the molecular level in terms of the dynamic process of reaching a chemical equilibrium. Any chemical system (all molecules within a certain volume) with a definable starting point will posses a certain chemical milieu, with each molecule having a certain energetic relationship to its neighbors based on its chemical bonds. As time moves forward, the molecules that are energetically disposed to persisting in this milieu will remain in their original molecular composition longer, and it is possible that some molecules may lead to the kind of self-replication that must have occurred here on earth at the beginning of life. Biochemists who investigate the chemical origins of life have applied the PNS to chemistry for decades now, but Rosenberg claims that the PNS properly understood is a basic law of physical science. If this turns out to be true, I see no reason to doubt that biological explanation will one day be reduced to biochemical explanation.


The second point is that I suspect the strong resistance to reductionism by biologists and philosophers is a result in large part to their fear of determinism. This is the only explanation I can come up with to understand how the majority of scientists and philosophers are both physicalists and non-reductionists at the same time. Rosenberg spends an entire chapter explaining how reductionism actually disproves genetic determinism, by which he means the idea that a single gene uniquely determines a single trait in an organism. This is true, genetic determinism popularly understood does not hold water, but reductionism certainly entails molecular determinism. If all of biology can be reduced to the motion of molecules in space, then where does free will fit into the picture? Because I find reductionism so convincing, I also find it hard to deny determinism, and this leaves me in a moral dilemma. If there is no free will, how can one say that a person aught to do one thing as opposed to another? That will have to be a topic for another essay.

2 comments:

silentio said...

Nate:

Interesting stuff that reductionism. I suppose the determinism explanation you end with doesn’t really help us understand anti-reductionism in biology (since if minds reduce deterministic interactions between cells then it doesn’t really matter that they don’t also reduce to interactions between molecules). But I suppose also that defending anti-reductionism in one realm could be thought to give support to the idea that anti-reductionism is a general phenomena that applies to all different “levels” of explanation (including psychological properties to neurophysiological properties).

I think it is important to note the different conceptions of reduction in the philosophy of science. There is the classic version which comes from Ernst Nagel requires that we have “bridge laws” which define a term from the reduced theory in terms of those of the reducing theory. Here the relationship is between theories or laws. This kind of reductionism isn’t really very much en vogue with philosophers of science today. The philosophy of science in the 1960’s with the question of incommensurability being raised put into question the whole scheme of reduction. Philosophers today have moved toward ideas of reduction which don’t require establishing identities or definitions between the terms of the different sciences (bridge laws).

I can understand the motivation for upholding a kind of explanatory pluralism. When what we want explained is something simple enough, what good are the other details? Putnam once made the same kind of point that Kitcher was making about chromosomes. Suppose we want to know why the square peg fits through the square hole but not through the round hole. Relative to that question, a microphysical explanation is unhelpful. Regarding ontological reduction, I can even understand the motivation for being a kind of pragmatist about what is real. None of this, of course, requires the rejection of basic physicalist premises (supervenience, token-identity). I understand, of course, that the reductionist wants to know about the causal influence of the higher levels. That’s where the real question lies.

Sub-sub-librarian said...

Silentio,

I think you meant to write “since if minds [don’t] reduce [to] deterministic interactions between cells then it doesn’t really matter that they don’t also reduce to interactions between molecules.” Correct?

I think you’re right to say that it’s important to note the different conceptions of reductionism (Nagel’s bridge laws, etc.). In this post, along with the following post which was a summary, I’m basically opting for a firm rejection of the extremes: a Kauffman type emergence and what Dennett called greedy reductionism. Kauffman argues for ontological antireductionism, and greedy reductionism is the insistence that our knowledge of molecules in motion will eventually explain something like World War II. What I think makes sense is the moderate zone between the extremes, where the assumption is physicalism and the issue is scientific explanation, and to prefer a reductionist program without denying the benefits of higher level investigations where people like complexity theorists do their stuff.

As for the issue of reductionism and the mind, I guess I’m not convinced that a thorough investigation of interactions between cells won’t one day be able to say something significant about consciousness. Can the mind be reduced to firing neurons, in the sense that consciousness is “nothing but” cellular interactions? Probably not. But reductionism is about explanation, and as you point out, what we want explained will dictate the level of details. I’m just convinced that the more details we collect the more interesting things we’ll be able to say by taking a reductive tact.