:: Article

From A Biological Point Of View, and Then Some

Interview by Richard Marshall.


Evolutionary theory, properly understood, does not conflict with the idea that God occasionally intervenes in nature – for example, by once or twice causing a beneficial mutation to occur. Biologists have not detected any such interventions despite the data and theory they have assembled about mutation. However, I think it is a mistake to expect biological experiments to be able to detect such one-off acts of divine intervention, especially if those acts occurred in the distant past. Science isn’t in that line of work.

By epistemically relevant, I mean that when one theory is simpler than its competitor, this fact is relevant to saying what the world is like. The issue is not whether simpler theories are beautiful, or easy to remember or test, or can be written down with less ink.

The main flaw in Nagel’s criticism of evolutionary theory is his demand that remarkable facts must have high probabilities, given the initial state of the universe. I think that the existence of Beethoven is remarkable, but I do not bristle at the suggestion that this event had a low probability given the initial state of the universe.

Elliott Sober is Hans Reichenbach Professor and William F. Vilas Research Professor in the Philosophy Department at the University of Wisconsin—Madison. here he discusses methodological naturalism, the relationship between evolutionary theory and theism, Ockam’s Razor and parsimony paradigms, parsimony’s relationship with instrumentalism, evolutionary theory and group selection, evidentiary relationship of common ancestory and natural selection, testing hypotheses about natural selection, what Fodor gets wrong, what Nagel gets wrong, whether physicalism is the default position of biology, supervenience, how biological science handles race and the Quine/Putnam indispensability argument. This covers the long ground and keeps on rollin’…


3:AM: What made you become a philosopher?

Elliott Sober: When I was in high school I found literature and history interesting, but science not at all. Literature and history obviously involved thinking, but science seemed to be all about memorizing facts and doing mindless calculations. When I was an undergraduate at University of Pennsylvania in the 1960’s during the heyday of Student Power, I was able to put my convictions into practice by not studying science. In the history and literature courses I took, epistemological questions came to interest me most. What makes one explanation of the French Revolution better than another? What makes one interpretation of “Waiting for Godot” better than another? These questions led me to philosophy and then to philosophy of science. By this back door, I finally got interested in science. However, I didn’t get into philosophy of biology until I started teaching at University of Wisconsin. I was intrigued by Bill Wimsatt’s review of George C. Williams’s Adaptation and Natural Selection, so I read the book. I was wowed.

3:AM: You’re a leading philosopher of science and biology. You’ve defended methodological naturalism, a species of naturalism, and you’ve made some interesting arguments about the relationship of evolutionary theory to theism. Can you sketch both?

ES: “Methodological naturalism” and “metaphysical naturalism” are terms that often surface in the continuing battle between evolutionary biology and creationism/intelligent design. The methodological thesis says that scientific theories shouldn’t postulate supernatural entities; the metaphysical thesis says that no such entities exist. In this debate, God is the supernatural entity at issue; the question isn’t whether science gets to talk about mathematical entities if Platonism is correct. Biologists often maintain that evolutionary theory abides by the 1st ism without embracing the 2nd. I agree with this interpretation despite what creationists and some atheists have maintained. I disagree with those who argue that evolutionary biology and the existence of God are incompatible.

This is not to deny that there are versions of theism that do conflict with evolutionary biology. Young Earth Creationism is an example; it claims that God created life on earth within the past 10,000 to 50,000 years. But other types of theism are different. Deism, for example, is compatible with evolutionary theory. Deism claims that God creates the universe and the laws of nature and then is hands-off, with everything that subsequently happens in nature being due to natural processes. Deism is logically compatible with evolutionary theory for the simple reason that the theory says nothing about the origin of the universe or of the laws of nature. More controversially, I also think that some “interventionist theisms” are compatible with evolutionary theory. (By “intervention,” I don’t mean that God violates laws of nature; I mean that God affects what happens in nature in ways that are additional to the ones that deism recognizes.) Evolutionary theory, properly understood, does not conflict with the idea that God occasionally intervenes in nature – for example, by once or twice causing a beneficial mutation to occur. Biologists have not detected any such interventions despite the data and theory they have assembled about mutation. However, I think it is a mistake to expect biological experiments to be able to detect such one-off acts of divine intervention, especially if those acts occurred in the distant past. Science isn’t in that line of work.

This point becomes more obvious if you consider the statement “guided mutations that are scientifically undetectable have occurred.” Many scientists will scoff at this statement, and I do too. However, the reason for scoffing is not that there is scientific evidence that such events never occur. Rather, the reason for scoffing is that there is absolutely no evidence that such events have happened. True, if you embrace the philosophical doctrine that whatever happens in nature is scientifically detectable, you will conclude that the statement in quotation marks is false. But now it is a philosophy (one that resembles the verificationism of the logical positivists), not a scientific theory, that is doing the talking.

I don’t endorse deism or interventionist theism. My point is just that evolutionary biology is logically compatible with the former and with some versions of the latter. I have bothered to make this point in print because I want to take the heat off of evolutionary biology. The more evolutionary theory gets called an atheistic theory, the greater the risk that it will lose its place in public school biology courses in the United States. If the theory is thought of in this way, one should not be surprised if a judge at some point decides that teaching evolutionary theory violates the Constitutional principle of neutrality with respect to religion. Creationists have long held that evolutionary theory is atheistic; defenders of the theory do the theory no favor when they agree.


3:AM: Are there other naturalisms that are philosophically interesting?

ES: Methodological naturalism, as I just described it, gives advice to scientists about what they should include in their theories. There is a second type of methodological naturalism that gives advice to philosophers, which I call “methodological naturalismp.” It says that the methods that philosophers should use in assessing philosophical theories are limited to the methods that scientists ought to use in assessing scientific theories. Methodological naturalismp has intrigued me in connection with the principle of parsimony. In my 2015 book, Ockham’s Razors – A User’s Manual, I devote a chapter to examining various parsimony arguments that philosophers have made. How well do these arguments measure up when judged by the standards that govern the parsimony arguments made in science? The answer is a mixed bag. Some philosophical arguments (e.g., in connection with the mind/body problem) look pretty good, while others (e.g., those that criticize moral realism) do not. By the way, I don’t assume that methodological naturalismp is true; rather, my work on philosophical parsimony arguments explores one context in which it might be.

3:AM: Speaking of Ockham’s Razor, the principle of parsimony seems at first blush no less arbitrary a principle than “complex is better” – so why is simpler better?

ES: I think there are three “parsimony paradigms” that each show how parsimony can be epistemically relevant. By epistemically relevant, I mean that when one theory is simpler than its competitor, this fact is relevant to saying what the world is like. The issue is not whether simpler theories are beautiful, or easy to remember or test, or can be written down with less ink.

The first parsimony paradigm connects parsimony and probability. Earlier attempts to show that simpler theories always have higher prior probabilities have failed, but there is a restricted circumstance in which the claim is right. If you have evidence that C1 is a cause of E, and no evidence as to whether C2 is also a cause of E, then C1 seems to be a better explanation of E than C1&C2 is, since C1 is more parsimonious. I call the version of Ockham’s razor used here “the razor of silence.” The better explanation of E is silent about C2; it does not deny that C2 was a cause. The problem changes if you consider two conjunctive hypotheses. Which is the better explanation of E, C1&notC2 or C1&C2? The razor of silence provides no guidance, but another razor, the razor of denial, does. It tells you to prefer the former explanation. The razor of silence is easy to justify; justifying the razor of denial is more difficult, so let me now turn to cases in which the competing hypotheses are incompatible with each other. This isn’t what is going on when you compare C1 and C1&C2.

One such case of incompatible competing hypotheses involves the comparison of common cause and separate cause explanations. Evolutionary biologists often appeal to parsimony when they seek to explain why organisms “match” with respect to a given trait. For example, why do almost all the organisms that are alive today on our planet use the same genetic code? If they share a common ancestor, the code could have evolved just once and then been inherited from the most recent common ancestor that present organisms share. On the other hand, if organisms in different species share no common ancestors, the code must have evolved repeatedly. The common ancestry hypothesis is more parsimonious, since it requires fewer changes to have occurred in lineages. Building on ideas developed by Hans Reichenbach, I’ve identified assumptions that entail that current organisms have a higher probability of sharing a single code if the common ancestry hypothesis is true than they’d have if the hypothesis of separate ancestry were true. That is, the simpler hypothesis has the higher likelihood in the technical sense of “likelihood” used in statistics. This is the second parsimony paradigm. It is an interesting consequence of this likelihood framework that common cause hypotheses don’t always have higher likelihoods. This is because there are nonReichenbachian assumptions that are sometimes plausible and that entail that a separate cause explanation of a similarity is more likely than a common cause explanation. “Simpler is always better” is an overstatement.

The third parsimony paradigm involves the assessment of scientific models for their predictive accuracy. The statistician Hirotugu Akaike proved a remarkable theorem in the 1970’s that demonstrates the relevance of parsimony to estimating a model’s predictive accuracy. Akaike’s work is part of a thriving branch of statistics called “model selection theory.” My colleague Malcolm Forster and I started working on model selection in the 1990’s.

These three paradigms describe three ways in which parsimony can be epistemically relevant. In many contexts, simplicity is not an aesthetic frill. I don’t claim that these paradigms provide a complete account of when parsimony arguments are justified. Still less do I claim that they sanction every parsimony argument that has been dreamed up. There are good parsimony arguments and bad ones. Maybe someday a unified account of parsimony will become available, but I don’t see much hope for that right now. My pluralism about parsimony paradigms is what led me to use the plural in the title of my book, Ockham’s Razors.

3:AM: Does the principle of parsimony mean that one is buying into instrumentalism from the get go?

ES: That depends on which of the three parsimony paradigms you’re thinking about. In the first and second, you are assessing hypotheses for their truth. In the first, you’re considering whether C1 or C1&C2 has the higher probability of being true. In the second, you’re considering common ancestry and separate ancestry and you’re wondering whether the shared genetic code favors one of those truth claims over the other. However, in the third paradigm, you’re aiming at predictive accuracy, not truth. It is an interesting fact about model selection that the evidence at hand can indicate that a model known to be false will be more predictively accurate than a model known to be true. This opens the door to a kind of instrumentalism. Scientists often seek predictively accurate models, rather than models that are true. Unfortunately, philosophers of science usually regard scientific realism and scientific anti-realism as monistic doctrines. The assumption is that there is one goal of all scientific inference – finding propositions that are true, or finding propositions that are predictively accurate. In fact, there are multiple goals. Sometimes realism is the right interpretation of a scientific problem, while at other times instrumentalism is.


3:AM: In addition to your work on common ancestry and parsimony, what other philosophical questions about evolutionary theory have you found intriguing?

ES: I have spent a lot of time arguing that the theory of group selection is not the stupid, pernicious doctrine that many biologists once claimed it to be. The theory is not just conceptually coherent; there are adaptations out there in nature (like reduced virulence in some viruses) that evolved because there was group selection. And our own species evolved under the influence of group selection, as Darwin emphasized when he discussed the evolution of altruism. Group selection and individual selection are just two of the selection processes that have played important roles in evolution. There also is selection within individual organisms (intragenomic conflict), and selection among multi-species communities (an idea that now is getting attention in work on the human microbiome). All four of these levels of selection find a place in multi-level selection theory.

The rabid opposition to group selection has now considerably subsided. In the process, the conceptual structure of evolutionary theory has become clearer, as have the relationships that connect different theoretical approaches. For example, evolutionary game theory was originally developed as an alternative to the hypothesis of group selection; now it is clear that game theory models postulate group selection, even if they do not use the g-word. David Sloan Wilson and I tell this historical and conceptual story in our 1998 book Unto Others – The Evolution and Psychology of Unselfish Behavior. As the title indicates, part of the book is about psychological egoism. We consider philosophical arguments and psychological experiments that have been taken to refute egoism. We are critical of both, but propose an evolutionary reason for thinking that psychological egoism is mistaken.

I also have been interested in how probability concepts are used in evolutionary theory. For example, how should the concept of fitness be understood? Another question I’ve worked on is whether the probability concepts used in the theory are “objective.” One influential philosophical position about the use of probability in science holds that probabilities are objective only if they are based on micro-physics; all other probabilities should be interpreted subjectively, as merely revealing our ignorance about physical details. I have argued against this position, contending that the objectivity of micro-physical probabilities entails the objectivity of macro-probabilities.


Another philosophical question about evolutionary theory that has interested me is the evidential relationship of common ancestry and natural selection. The two are logically independent, but this leaves open how evidence for one is related to evidence for the other. Darwin repeatedly used the hypothesis of common ancestry as a platform on which to build his various ideas about testing hypotheses concerning natural selection. He also argued that adaptive similarities provide little or no evidence for common ancestry. Although this second claim needs to be fine-tuned, Darwin was right that ample evidence for common ancestry can exist even if none of the characteristics we observe were caused to evolve by natural selection. The big picture, I think, is that common ancestry is evidentially prior to natural selection in Darwin’s theory and in contemporary evolutionary biology as well. This is the main point of the first chapter of my 2011 book Did Darwin write the Origin Backwards?

3:AM: Speaking of the task of testing hypotheses about natural selection, how does evolutionary biology do this?

ES: If the organisms in a species now have trait T, and this trait now helps those organisms to survive and reproduce because the trait has effect E, a natural hypothesis to consider is that T evolved in the lineage leading to those current organisms because T had effect E. This hypothesis is “natural,” but it often isn’t true! Darwin and his successors have railed against the fallacy of confusing the current utility of a trait with the reason the trait evolved. For example, Darwin argued that skull sutures in mammals did not evolve because they facilitate live birth; the sutures were in place well before live birth evolved. Checking the chronological order in which different traits evolved in a lineage is one way to test an adaptive hypothesis; the fact of common ancestry is what makes that checking possible.

Another way to test hypotheses about adaptation is to consider trait variation across a group of species instead of focusing on the trait of a single species. Rather than seeking to explain why polar bears have fur of a certain thickness, one tries to explain why bears in colder climates have thicker fur than bears in warmer climates. The former problem is hard to solve, since it is hard to say exactly what fur thickness polar bears should have if natural selection guided the evolution of that trait. The latter problem can be solved without one’s having to say what the optimal fur thickness is for polar bears. The adaptive hypothesis predicts a negative correlation between fur thickness and ambient temperature, and this can be tested against an alternative hypothesis that says that fur thickness should be uncorrelated with ambient temperature. The fact that bear species have a common ancestor helps make this argument work, a point that Steve Orzack and I have defended.

3:AM: Jerry Fodor, both as single author and in collaboration with Massimo Piattelli-Palmarini, has argued that the theory of natural selection is in-principle incapable of explaining adaptations. Does Fodor have a point? What did he get wrong?

ES: Fodor thinks that the theory of natural selection can’t explain the adaptive features of organisms because he thinks there can be no laws of natural selection. There are no laws, he says, because “who wins a … competition [between two traits] is massively context sensitive.” Fodor’s complaint is not that the laws have ceteris paribus clauses; for Fodor, it is okay for special science laws to have an “all else being equal” proviso. Fodor’s beef is that “it simply isn’t true, for example, that being big is in general better for fitness than being small except when there are effects of interacting variables; or that flying slow and high is in general better for fitness than flying fast and low except when there are effects of interacting variables; or that being monogamous is in general better for fitness than being polygamous … except when there are effects of interacting variables … It’s not that the underlying generalizations are there but imperceptible in the ambient noise. It’s rather that there’s just nothing to choose between (e.g.) the generalization that being small is better for fitness than being big and the generalization that being big is better for fitness than being small.” For Fodor, “the theory of natural selection reduces to a banal truth: if a kind of creature flourishes in a kind of situation, then there must be something about such creatures (or about such situations, or about both) in virtue of which it does so.”


One of Fodor’s mistakes is the idea that if there are laws of natural selection, they must have the simple form he describes. A better format would be this: trait X is fitter than trait Y in a population of organisms if those organisms have other biological traits T and live in an environment that has properties E. The theory of natural selection is filled with statements of this form. In my 1984 book The Nature of Selection, I call laws of this kind “source laws.” They describe synchronic supervenience bases for variation in fitness. They contrast with “consequence laws,” which describe how a population will change if there is variation in fitness.

3:AM: Thomas Nagel’s Mind and Cosmos caused a stir too as he argued against a materialistic reductionism. Where did he go wrong?

ES: Nagel thinks that “remarkable facts” can’t have low probabilities, given the initial state of the universe. He thinks that the existence of life, the existence of mind, and the existence of consciousness (call this triplet LM&C) are all remarkable. He further thinks that contemporary evolutionary theory says that LM&C had very low probabilities of occurring, so the theory must be false or seriously incomplete. Nagel calls for the construction of a new, teleological science, one that shows that LM&C were in the cards from the universe’s first moment. According to this sought-for theory, earlier events happen because they will help LM&C to appear later. The teleology that Nagel wants isn’t theistic, and it isn’t causal; he doesn’t think that later events cause earlier ones.

The main flaw in Nagel’s criticism of evolutionary theory is his demand that remarkable facts must have high probabilities, given the initial state of the universe. I think that the existence of Beethoven is remarkable, but I do not bristle at the suggestion that this event had a low probability given the initial state of the universe.

Another problem with Nagel’s position is that it isn’t clear what probabilities current science assigns to the existence of LM&C, given the state of the universe at the time of the Big Bang. Part of the problem is the absence of a detailed conception of what needs to happen in the interval between the Big Bang and the emergence of LM&C. There’s also a problem concerning what the propositions are whose probabilities we’d like to assess. Is it the fact that LM&C each exist somewhere in the universe at some time or other? Or is it the fact that LM&C all appeared on Earth within the past 4 billion years? If the former, maybe the probability isn’t so low. If the latter, then maybe the probability is kind of low. Notice that I just said “maybe” twice.

3:AM: Is physicalism the default position of biology and if it is, how do the biological properties of a biological system relate to its physical properties? This issue raises what talk of supervenience means in this context – and might it be helpful in getting difficulties of, say, mind/body problems solved?

ES: Biologists now pretty universally regard vitalism as a vestige of a bygone age. And philosophers of biology generally recognize that evolutionary fitness (roughly, an organism’s ability to survive and reproduce in its environment) is multiply realizable. The upshot is that most philosophers of biology now hold that biological properties supervene on physical properties (where supervenience is taken to include some kind of “in virtue of” relation), and that fitness and other biological properties are not identical with physical properties.


I am inclined to think that vitalism is related to physicalism about biological properties in the same way that dualism is related to physicalism about mentalistic properties. Just as thought experiments can’t show that vitalism is true (or that it is false), they also can’t show that dualism is true (or that it is false). The same point holds for supervenience claims. This is because I think that the meanings of biological and psychological concepts leave open which of these philosophical theses is true. It can be a necessary conceptual truth that pains are painful without this ruling out the physicalist thesis that immaterial minds are impossible or the thesis that conscious states supervene on physical states. The necessity involved in these claims is nomological necessity, not metaphysical necessity (assuming that these are different). I disagree with the widely held view that it is metaphysical necessity, not nomological, that matters in the mind/body problem.


The sort of “semantic openness” I just described for biological and psychological predicates is, I think, characteristic of theoretical terms in science. An interesting example of semantic openness from philosophy of biology has to do with the concept of species. Philosophers have often held up biological species as paradigmatic examples of natural kinds, but Michael Ghiselin and David Hull made a good case for thinking that species are historical particulars, not natural kinds. “Tiger is a natural kind” and “Tiger is a historical particular” are incompatible with each other, and evolutionary biology provides a reason for favoring the latter over the former. I take this to mean that the definition of the term “tiger” doesn’t settle which ontological category the tiger species belongs to.

3:AM: How does biological science handle race?

ES: Evolutionary biologists often avoid using the term “race” because there is so much racist baggage that comes with the term. However, they are often okay with the idea that the genealogy of human groups within our species can sometimes be inferred in much the same way as the genealogy of different species. Sometimes the inferred genealogies indicate that the racial categories that are used in a given society (for example, in contemporary America) are biologically meaningless, but sometimes it turns out that a vernacular racial category has biological reality. Robin Andreasen has explored the philosophical ramifications of this biological work.

The situation in biomedicine is different. Instead of thinking of the question of race genealogically, and leaving it open whether vernacular races are genealogical units, the interest in biomedicine has been to determine whether vernacular racial categories are medically useful in diagnosis and treatment. There is on-going debate about this.

3:AM: You have been critical of the Quine/Putnam indispensability argument for the existence of mathematical entities. Can you describe what your criticism is?

ES: The indispensability argument says (roughly) that if you have ample reason to accept an empirical scientific theory that makes indispensable use of mathematics, and that theory entails that numbers exist, then you have ample reason to accept that numbers exist. The argument affirms the antecedent of this conditional, and concludes that you have ample reason to believe that numbers exist. What is striking about this argument is that it seems to show that the empirical reasons that suffice for accepting a scientific theory also suffice for accepting a metaphysical claim.

The conditional at the heart of this argument looks fine if accepting a proposition is just a matter of its being sufficiently probable ─ e.g., having a probability greater than 0.95. After all, if T has a probability greater than 0.95, and T entails M, then M also must have a probability greater than 0.95. However, the indispensability argument is usually understood as a claim about evidence – that the empirical evidence you have for a scientific theory is also evidence that numbers exist. It is this evidence claim that leads the indispensability argument astray. To see why, consider the following analogy.

Suppose you are playing poker and you wonder whether the card you are about to be dealt will be the Jack of Hearts. The dealer is a bit careless and so you catch a glimpse of the card at the top of the deck before it is dealt to you. You see that it is red. The fact that the card is red is evidence that the card will be the Jack of Hearts, and the hypothesis that the card will be the Jack of Hearts entails that the card will be a Jack. However, the fact that the card is red isn’t evidence that the card will be a Jack. From the fact that E is evidence for T and the fact that T entails M, it doesn’t follow that E is evidence for M. This point about evidence has been known for a long time. Carl Hempel wrote about “the special consequence condition of confirmation” in the 1960’s, and it didn’t take long for philosophers of science to come up with counterexamples like the one I just described.

The indispensability argument seeks to assimilate the epistemology of metaphysical statements to the epistemology of statements that are obviously empirical. I think it fails to achieve this goal. The argument does not refute the Carnapian thesis that scientific theories and metaphysical claims differ epistemologically – observations can provide evidence for the former, but not for the latter. Not that I subscribe to that Carnapian thesis, but that is another story …

3:AM: And for the readers here at 3:AM, are there five books other than your own that you can recommend to take them further into your philosophical world?

ES: 9780387953649

ES: Kenneth Burnham and David Anderson, Model Selection and Multimodel Inference ─ A Practical Information-Theoretic Approach (2nd edition).


Charles Darwin, On the Origin of Species by Means of Natural Selection.


Michael Friedman, Reconsidering Logical Positivism.


Ian Hacking, The Logic of Statistical Inference.


George C. Williams, Adaptation and Natural Selection.


Richard Marshall is still biding his time.

Buy his book here to keep him biding!

First published in 3:AM Magazine: Sunday, October 9th, 2016.